2.3 Modeling with Linear Functions

Turns a described situation into a linear rule: choosing which quantity is the input, reading the rate and the initial value out of the words, and — the step most often missed — stating the domain the situation allows rather than the one the formula would accept. Distinguishes interpolation from extrapolation and names model breakdown as the reason the distinction matters.

Subject: Precalculus · 69 slides · symbolic lesson

Open the interactive version of this deck

What this lesson covers

The lesson, slide by slide

1. Modeling with Linear Functions

Title

Precalculus · Section 2.3

Turning a paragraph into a rule, and knowing where the rule stops

2. What this lesson gives you

Objectives

Sections 2.1 and 2.2 gave you the algebra of a straight line: slope, intercepts, and the two forms of the equation. This lesson gives you the paragraph that comes before the algebra, and the sentence that has to come after it.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 232-245 — learning objectives 2.3.1 to 2.3.3

Nothing in this lesson is new algebra. You already know how to find a slope from two points, how to read an intercept, and how to write a line in slope-intercept form. What is new is everything that surrounds the algebra: turning a paragraph of English into those two numbers, and then turning the answer back into a sentence about the situation.

That sounds soft, and it is not. Most of the marks lost on this material are lost outside the algebra: a rate written positive when the quantity is falling, an intercept computed for a year two thousand years ago, an answer of minus fifteen months reported as though it meant something. Each of those is a failure to keep track of what the symbols stand for.

The last two objectives are the ones people skip. A model carries a domain with it, and a prediction made far outside the data it was built from is a different kind of claim from one made between two data points. By the end of this lesson you should be uncomfortable writing a model without also writing where it is allowed to be used.

3. Before anything new: a line from two points

Warm-up

Discussion prompt

A line passes through the points (0, 40) and (8, 64). Without looking back at Section 2.1, write its slope, its vertical intercept, and its equation in slope-intercept form.

Write your three answers down before revealing them. If any of the three took you more than a few seconds, go back to Section 2.1 before going on, because every worked example in this lesson ends with exactly this calculation buried inside it.

Notice that the second point did almost no work. Because the first point was at input zero, it handed you the intercept directly, and the second point was only needed for the slope. That accident is something you will deliberately arrange later in this lesson by choosing where your input starts counting.

Keep an eye on what the slope of 3 would mean if these were real quantities. If the input were weeks and the output dollars, the slope would be 3 dollars per week, and the intercept 40 dollars at the start. The algebra does not change; the sentence attached to it does.

4. From a paragraph to a rule

Section

Part 1

5. A model is a formula, a pair of units, and a domain

Concept

A linear model is the familiar rule with three extra pieces of information attached: what the input measures, what the output measures, and which inputs the situation actually permits.

\[ f(x) = mx + b \]

linear model — A linear function used to describe a real situation, together with the meaning and units of its input and output and the set of inputs the situation allows.

Strip those away and you have algebra with no subject. A question that asks you to interpret the slope is asking about the units; a question that asks for a reasonable domain is asking about the situation, not about the formula.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 232-233 — identifying steps to model and solve problems

The definition here is doing more work than it looks. Two students can write the same three symbols and only one of them has a model, because only one can say what the input measures, in what units, and over what stretch the rule is being claimed.

Think of it as a promise. When you write a model you are promising that for these inputs, in these units, the output is what really happens. Nothing in the algebra makes that promise; you make it, and you are responsible for stating its limits.

This is also why exam questions on this section so often ask you to interpret rather than to compute. Interpreting the slope means giving its units and saying what one extra unit of input does to the output. Interpreting the intercept means saying what is true before anything has happened. Both are answered in words, and both are worth marks.

6. Naming the two changing quantities

Concept

Emily has saved 3,500 dollars for a summer in Seattle and expects to spend 400 dollars a week on rent, food and activities. Two quantities change: time, and money.

The dependence runs one way. How long she has been there decides how much money is left; the money left does not decide the date. That is the test to apply whenever the choice is not obvious.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 233 — building linear models

Get into the habit of writing these three lines out before anything else, exactly as they appear here: output with its letter and units, input with its letter and units, and the statement that the output is a function of the input. It takes fifteen seconds and it prevents most of the errors in this lesson.

The test for which quantity is the input is the direction of dependence. Ask yourself which one you could decide freely, and which one follows. Emily can choose how long to stay; the money left then follows. She cannot choose to have 900 dollars left and thereby fix the date.

There is one more reason to choose carefully, and it is practical. Whichever quantity you make the output is the one you can compute by substitution; the other one costs you an equation to solve. So if the question ends with how much money, make money the output.

7. Two numbers describe every week of the summer

Intuition

Once the savings and the weekly spending are in hand, you are no longer limited to the two facts the paragraph gave you. You can answer about any week at all, including weeks nobody mentioned.

\[ M(3) = 2300, \qquad M(5) = 1500, \qquad M(7) = 700 \]

That is the leverage a model buys. Two numbers, read out of one sentence, stand in for a table of every week of the trip, and they also answer the reverse question: which week leaves her with 1,500 dollars.

It is also where the responsibility comes from. A rule that answers questions about weeks nobody asked about will just as happily answer about week thirty, where it describes nothing at all.

This slide is about what you have gained by building the model at all. The paragraph told you two things: how much she saved and how much she spends. From those two numbers you can now produce the balance in any week whatsoever, without being told anything more.

Check one of the three values by hand to see how little work it takes. After five weeks she has spent five lots of 400, which is 2,000, leaving 1,500 of the 3,500. That is the entire computation, and the model just packages it so that you can do it for any week at once.

It works in reverse too, which is the part worth noticing. Ask which week leaves her with 1,500 dollars and you are solving rather than substituting, but it is the same two numbers doing the work. Every question in this lesson is one of those two directions.

The warning in the last line is the theme of the second half of the lesson. A rule that will answer about weeks nobody asked about will also answer about week thirty, long after the money has gone, and it will do so with a straight face.

8. Reading Emily's model piece by piece

Notation

Annotate

On: \( M(t) = -400t + 3500 \)

  • The output: dollars remaining, after t weeks. The letter is chosen to remind you what it measures, and the units travel with it.
  • The rate of change, in dollars per week. Output units divided by input units, which is always what a slope measures.
  • It comes from the verb 'spending', not from any number in the sentence. A decreasing quantity has a negative slope.
  • The initial value: the output when the input is zero, before any week has passed. On a graph it is the vertical intercept.
  • Weeks since the trip began. Not the date, not her age: a count of weeks starting at zero, chosen so that the intercept is a number the problem gave.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 233 — initial value and rate of change

Step through the annotations in order and connect each to the sentence it came from. Only two numbers appear in the paragraph about Emily, and each one shows up in exactly one place in the model, so there is very little to remember and quite a lot to keep straight.

The third annotation is the one worth dwelling on. The paragraph never contains a negative number; it contains the word spending. The sign of a rate is carried by the verb, and every model you build for the rest of this lesson will need you to read the verb and choose the sign yourself.

The last annotation looks like a technicality and is not. Because the input counts weeks from the start of the trip, input zero is the moment she arrives with her savings intact, which is why 3,500 is the intercept. Had you counted weeks from her birthday, the same situation would have a completely different intercept and the same slope.

9. Emily in Seattle: writing the model

Worked example

Write a linear model for the money Emily has left after t weeks.

Name the input and the output, with units

Why: Money left depends on the length of the stay, so time is the input.

\[ t = \text{weeks since the trip began}, \quad M = \text{dollars remaining} \]

Read the initial value out of the words

Why: What she has before any week has passed.

\[ M(0) = 3500 \]

Read the rate of change out of the words

Why: Spending 400 a week lowers the output by 400 for each unit of input.

\[ m = -400 \text{ dollars per week} \]

Start from the slope-intercept form

Why: The rate is constant, so a linear function is the right shape.

\[ M(t) = mt + b \]

Substitute the rate and the initial value

Why: Nothing else has to be found.

\[ M(t) = -400t + 3500 \]

Figure (svg): The line M equals 3500 minus 400t drawn from t equals 0 to t equals 10 weeks, with the region under it from 0 to 8.75 shaded, the starting value 3,500 marked on the vertical axis and the crossing at 8.75 weeks marked on the horizontal axis.

Every week the line drops the same 400 dollars. The shaded stretch is the only part of the line that describes Emily's summer.

Check the units and one value

Why: Dollars per week times weeks gives dollars, so the units agree; after two weeks she should have spent 800.

\[ M(2) = -400(2) + 3500 = 2700 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 233 — the Seattle model

This is the model solution for the whole lesson: name, read, read, form, substitute, check. Every worked example that follows is this template with harder numbers, so it is worth being able to reproduce the five step headings from memory.

The fourth step, writing the slope-intercept form before substituting anything, feels like a wasted line. It is not. It forces you to decide that a linear function is the right shape, and that decision rests on the phrase each week: a constant amount per unit of time is exactly what a constant rate of change means.

Look at the check. Two different things are being verified, and both are cheap. The units multiply correctly, dollars per week times weeks giving dollars, which would catch a rate written upside down. And one concrete value is tested against common sense: after two weeks of spending 400 a week she should be 800 down, and she is.

10. Emily in Seattle: when does the money run out?

Worked example

Find the horizontal intercept of the model, and say what it means.

Translate the question into an equation

Why: Out of money means the output is zero.

\[ 0 = -400t + 3500 \]

Move the rate term to the other side

Why: Add 400t to both sides.

\[ 400t = 3500 \]

Divide by the rate

Why: Both sides by 400.

\[ t = \frac{3500}{400} = 8.75 \]

Say what the number means, with its unit

Why: The input was measured in weeks.

\[ t = 8.75 \text{ weeks} \]

State the domain the situation allows

Why: Before the trip there is no model, and after the money is gone there is no model either.

\[ 0 \le t \le 8.75 \]

Figure (svg): Stems and dots showing the money remaining at the end of weeks 0 through 10: 3,500 down to 300 at week 8, then negative values at weeks 9 and 10 drawn in red below the axis.

The last two dots are what the formula computes and the situation forbids. The model stops where the money does.

Check the intercept and one week before it

Why: At 8.75 weeks the balance is zero, and a week earlier there should still be a few hundred dollars.

\[ M(8.75) = 0, \qquad M(8) = 300 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 233-234 — the horizontal intercept and the reasonable domain

The phrase to translate here is out of money. Almost every modelling question hides an equation inside an English phrase like this one, and the translation is the step that earns the mark: out of money means the output is zero, which means you are looking for the horizontal intercept.

Notice which kind of question this is. You were given an output and asked for an input, so this is a solve, not an evaluation. Keeping those two apart in your head is most of the skill in this section, because the algebra needed is completely different even though the model is the same.

The last two lines are where the modelling happens. The number 8.75 is the answer to the algebra; the sentence that the money lasts eight and three quarter weeks is the answer to the question; and the domain line records the fact that the model describes nothing before week zero or after week 8.75. Look at the red dots on the figure: those are the weeks the formula is happy to compute and the situation cannot support.

11. Before any algebra: what is being asked for?

Step zero

A paragraph describes a situation and ends with a question.

Discussion prompt

Name the two things you decide before writing a single equation, and say why the order matters.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 232-233 — the eight modelling steps

Write your two answers before revealing. This slide exists because the most expensive mistakes in modelling problems are made in the first thirty seconds, before any algebra has been written down.

Reading the question first and making that quantity the output sounds like a trick, and it is a completely legitimate one. It converts the final step from solving an equation into a substitution, which is faster and much harder to get wrong. When the question asks for two different things, as several exercises here do, pick the output that makes the harder one easy.

The second decision, where input zero sits, is free and it is yours. You will use it deliberately in the very next part of the lesson to make one of the given data points into the intercept, which saves you a whole calculation.

12. Which quantity is the input?

Sorting

Sort into buckets

In each pair of quantities from this section, which one is the input?

the first one listed
weeks in Seattle, and money remaining; doughnuts produced, and total cost; miles driven, and the rental bill; hours walked, and miles from the intersection; months of payments, and loan balance; megabytes used, and the monthly phone bill
first
In every one of these the first quantity is the input: it is the one you choose or count off, and the second is what follows from it. Cost follows from production, the bill follows from the miles, the balance follows from the payments.
second
None belongs here, and that is the point. Nothing in this list has a bill deciding how far you drove, or a balance deciding how many months have passed.

Sort all six before checking. They are all from this section, so if one of them is unfamiliar, that is a sign to look back rather than to guess.

Every one of them belongs in the same bucket, which is the point of the exercise: word problems are almost always phrased with the driving quantity first. But do not let that become the rule you use. The real test is the direction of dependence, and it is worth saying the sentence out loud in both directions to hear which one is nonsense.

One of the six is worth a second look. Miles driven and the rental bill could conceivably be reversed if someone asked how far you can drive on a fixed budget. That question would make the bill the input, and the slope would then be miles per dollar rather than dollars per mile. Both models are correct; the question decides which one you want.

13. Swap the input and the output

Prediction

\[ M(t) = -400t + 3500 \]

Predict first

Suppose instead we take the money already spent as the input and the number of weeks as the output. What is the new rate of change?

  • one four-hundredth of a week per dollar
  • −400 dollars per week
  • 400 weeks per dollar
  • unchanged at −400

Correct: one four-hundredth of a week per dollar

Why: Swapping the roles turns the rate upside down, and the units turn upside down with it: dollars per week becomes weeks per dollar. Spending 400 dollars takes one week, so one dollar takes one four-hundredth of a week, which is 0.0025. Both models describe the same summer; the second one answers questions about money directly and questions about time indirectly.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 233 — matching the units of the rate to the variables

Commit to one option before revealing. This is the most common conceptual question asked about modelling, and it takes ten seconds if you think in units and a long time if you do not.

Here is the ten-second route. The original slope is dollars per week. Swapping the input and the output must turn the units upside down, so the new slope is measured in weeks per dollar, and that immediately rules out two of the four options. The remaining question is whether it is 400 or one four-hundredth, and since spending 400 dollars takes one whole week, one dollar takes one four-hundredth of a week.

Neither model is the right one in any absolute sense. They describe the same summer and they are inverse functions of each other in the sense of Section 1.7. The one you want is whichever makes the question you were asked into a substitution.

14. Rate and initial value from words

Section

Part 2

15. The rate of change is a change divided by a change

Concept

Whatever the situation, the slope is built the same way: how much the output moved, divided by how much the input moved.

\[ m = \frac{\text{change in output}}{\text{change in input}} \]

Two things come out of that fraction and both are marked. The units of the slope are the output units divided by the input units, so a slope of 380 is never just 380: it is 380 people per year. And the sign is carried by the verb of the sentence, not by any number in it.

the sentence saysthe slope is
grows by 2,500 a yearpositive: 2,500 people per year
spends 400 each weeknegative: 400 dollars per week, going out
decreased by 205 each yearnegative: 205 people per year
costs 0.25 to produce each onepositive: 0.25 dollars per doughnut

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 235 — the change in output over the change in input

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 242 — Exercises 156 and 175

Everything in this lesson about rates comes out of that one fraction. When you are given two data points you compute it; when you are given the rate in words you read it. Either way the quantity on top is a change in the output and the quantity underneath is a change in the input, and getting them the wrong way round is the single most common slip in this section.

The units are your protection against that slip. A slope is never just a number here. It is people per year, dollars per week, dollars per megabyte. If you can say the units out loud and they sound like an answer to the question, you have the fraction the right way up.

Look at the table and read the verbs: grows, spends, decreased, costs. Two of them describe a quantity going down, and those two take a negative sign even though every number printed in the sentence is positive. Nobody will ever write the minus sign for you.

16. The initial value is the output at input zero

Concept

The initial value is whatever the output is when the input is zero. Sometimes the problem hands it to you as a starting amount or an up-front fee; sometimes it has to be computed from a point that is not at zero.

\[ b = f(0) \]

Hannah plans to pay off a no-interest loan from her parents. Her balance is 1,000 dollars and she pays 250 dollars a month until the balance is zero. The 1,000 is the value at zero months, and the 250 is the rate.

\[ f(x) = -250x + 1000 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 233-234 — using a given intercept to build a model

Hold on to the definition rather than the picture. The initial value is the output when the input is zero, and only on a graph is it also called the vertical intercept. That distinction matters in a moment, when you meet problems where input zero is not one of the given data points and the initial value has to be computed.

Hannah's loan is the easy case, and it is worth seeing exactly why. The problem hands you both numbers directly: a balance before any payment, which is the value at input zero, and a fixed monthly payment, which is the rate. There is nothing to compute before writing the model down.

As you read the next slide, notice that the sign of the rate again comes from a verb. She pays off the loan, so the balance falls, so the rate is negative. The number 250 appears in the sentence with no sign at all.

17. Hannah's loan: a model from an intercept and a rate

Worked example

Write the model and find how long the loan takes to clear.

Name the input and the output

Why: The balance depends on how many payments have been made.

\[ x = \text{months of payments}, \quad f(x) = \text{balance in dollars} \]

Read the vertical intercept

Why: The debt before any payment.

\[ b = 1000 \]

Read the rate, with its sign

Why: Paying reduces the balance.

\[ m = -250 \text{ dollars per month} \]

Write the model

Why: Slope-intercept form, since the value at zero is known.

\[ f(x) = -250x + 1000 \]

Set the balance to zero

Why: A cleared loan is a balance of zero.

\[ 0 = -250x + 1000 \]

Solve for the number of months

Why: Add 250x to both sides, then divide.

\[ 1000 = 250x \;\Longrightarrow\; x = 4 \]

Figure (svg): The line f of x equals 1000 minus 250x from x equals 0 to 5, with dots at months 0, 1, 2, 3 and 4 showing balances 1000, 750, 500, 250 and 0.

Four equal steps down from 1,000 land exactly on zero, which is why the horizontal intercept answers the question the problem asked.

Check with plain arithmetic

Why: Four payments of 250 dollars come to exactly the size of the loan, so the model and common sense agree.

\[ 4 \cdot 250 = 1000 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 234 — Hannah's loan

This example is short enough that you can check every step against your intuition, which makes it a good one to use as a template. Name, read the intercept, read the rate, write the model, translate the question, solve.

The translation step is the same one as in Emily's problem: cleared the loan means the output is zero. If you find that phrase hard to spot in a longer problem, try asking what number the output must equal, and see whether the problem supplies it directly or hides it in a phrase such as empty, paid off, sold out, or out of contact.

The check is deliberately not algebraic. Four payments of 250 dollars come to 1,000 dollars, which is the whole loan, so four months is obviously right. Whenever a check can be done by arithmetic that a ten-year-old could follow, do it that way, because it is independent of the algebra you are checking.

18. Match the phrase to the constant it gives

Matching

Match the pairs

  • l1. an up-front fee of 20 dollars
  • l2. then 59 cents a mile
  • l3. she saved 3,500 for the trip
  • l4. spending 400 each week
  • l5. a fixed cost of 25,000
  • r1. initial value 20
  • r2. slope 0.59 dollars per mile
  • r3. initial value 3,500
  • r4. slope −400 dollars per week
  • r5. initial value 25,000

Why: A phrase gives a rate when it contains the word per, or a unit divided by another unit hidden inside words like 'a mile' or 'each week'. A phrase gives an initial value when it describes an amount that exists before the input starts counting: a fee, a saving, a fixed cost, a starting balance.

Work through all five before checking. The skill being practised is reading, not algebra, and it is worth naming the two signals explicitly.

A phrase gives you a rate when it contains a per, either written out or hidden inside a preposition: 59 cents a mile, 400 each week, 0.25 to produce each one. A phrase gives you an initial value when it names an amount that exists before the input has started moving: a fee, a saving, a fixed cost, a balance.

The fourth pair is the one that catches people. Spending 400 each week is a rate, and it is negative, because the money it refers to is leaving. If you paired it with an initial value you were probably reading the number rather than the phrase around it.

19. Try It 2.14: the doughnut company's cost

Worked example

A company sells doughnuts. It has a fixed cost of 25,000 dollars for rent, insurance and other expenses, and each doughnut costs 0.25 dollars to produce. Write the cost as a function of the number produced, and interpret the vertical intercept.

Name the input and the output

Why: The total cost depends on how many are made.

\[ x = \text{doughnuts produced}, \quad C = \text{total cost in dollars} \]

The fixed cost is the value at zero

Why: It is paid whether or not a single doughnut is made.

\[ C(0) = 25000 \]

The cost per doughnut is the rate

Why: Dollars per doughnut is output units over input units.

\[ m = 0.25 \text{ dollars per doughnut} \]

Write the model

Why: Rate times input, plus initial value.

\[ C(x) = 0.25x + 25000 \]

Interpret the vertical intercept in the situation's own words

Why: Produce nothing and the rent is still due.

\[ C(0) = 25000 \text{ dollars with no doughnuts made} \]

Figure (svg): The line C of x equals 0.25x plus 25000 drawn for up to 40,000 doughnuts, starting at 25,000 dollars on the vertical axis and rising to 35,000.

The intercept towers over the slope: the first doughnut costs 25,000 dollars and every one after it costs a quarter.

Check with a production run

Why: Ten thousand doughnuts add 2,500 dollars of ingredients to the fixed 25,000.

\[ C(10000) = 0.25(10000) + 25000 = 27500 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 235 — Try It 2.14

Try this one on paper before reading the steps; it is the book's own practice problem and it is a close copy of Hannah's loan with the sign flipped. The fixed cost is the value at zero production, and the cost per doughnut is the rate.

The interpretation step is the part the question actually asks for, so do not skip past it. The vertical intercept is 25,000 dollars, and in the company's own words that is what the month costs if not a single doughnut is made: the rent, the insurance and the other fixed expenses are due either way.

Look at the shape of the graph. The line looks almost flat, because a quarter of a dollar per doughnut is tiny compared with a fixed cost of 25,000. That is worth knowing about real cost models: for small production runs the fixed cost dominates completely, which is why the average cost per doughnut falls so sharply as production grows.

20. Trap: taking the sign from the number, not the verb

Trap

The trap

Emily spends 400 dollars a week, so the model is written:

\[ M(t) = 400t + 3500 \]

Wrong. Nothing in the sentence carried a minus sign, so none was written.

The fix

The sentence says spending. Money leaves, so the output falls as the input grows, and a falling output means a negative slope. Read the verb, then choose the sign.

\[ M(t) = -400t + 3500 \]

The wrong model has Emily richer every week, and its graph rises. Sketching the line, even roughly, catches this in two seconds.

This is the single most common error in the whole section, and it is not a careless one: the sentence genuinely contains no negative numbers, so writing none feels right. The defence is to make the sign a separate decision, taken from the verb, every single time.

Build the habit of asking one question after writing any rate: as the input grows, does the output go up or down? Spending, paying off, draining, depleting, decreasing and losing all mean down. Saving, growing, gaining, charging per unit and producing all mean up.

A rough sketch catches it too. The wrong model rises, which would mean Emily has more money every week she stays in Seattle. Ten seconds with a pencil is enough to see that, and a sketch is worth making on any modelling problem you are unsure of.

21. Fill in the gym membership model

Fill the middle

A gym charges 60 dollars to join, then 35 dollars a month.

Fill in the blanks

C(m) = 35m + 60

Why: The monthly charge is a rate, so it multiplies the input; the joining fee is paid once and is therefore the value at zero months. Check it: at m equal to 0 the cost is 60, the joining fee alone, and each extra month adds 35. Writing them the other way round would say the gym charges 35 to join and 60 a month.

Fill both blanks before checking. This is the same reading skill as the matching slide, now with the two numbers competing for the same two slots, which is where the mistake actually happens.

The joining fee is paid once, at the moment you join, which is input zero. So it cannot multiply the input; it stands alone as the initial value. The monthly charge repeats with every month, so it must multiply the number of months.

There is a quick test that works on any model of this kind. Put zero in for the input and see what you get. A correct model returns the joining fee alone, because at zero months you have joined and paid nothing else. If putting zero in returns the monthly charge, your two numbers are swapped.

22. Two data points, one line

Section

Part 3

23. Two input-output pairs are enough

Concept

Often nobody tells you the rate. Instead you are given two moments: a population in one year and in another, two phone bills, two house values. Two pairs determine the line completely.

  1. Identify the input and output values
  2. Convert the data to two coordinate pairs
  3. Find the slope
  4. Write the linear model
  5. Use the model to predict, by evaluating it at a given input
  6. Use the model to find an input that produces a given output
  7. Answer the question that was posed

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 234 — using a given input and output to build a model

This is the situation in most exercises and in almost all real data: nobody tells you the rate, they tell you two measurements. Two points determine a line exactly, which is why two measurements are enough to fix the whole model.

Read the seven steps as three groups. The first two are about setting up, and they are where you choose the units and where input zero sits. The middle three are the algebra from Section 2.1. The last two are about answering, and notice that the list distinguishes them: predicting means evaluating at an input, while finding the input that produces a given output means solving.

The very last step, answer the question posed, is there because the algebra does not do it for you. A number with no units and no sentence around it is only half an answer, and in a word problem it is the half worth fewer marks.

24. Choose where the input starts

Concept

You are free to decide what input zero means, and the choice is worth a minute's thought. Using the calendar year as the input puts the intercept more than two thousand years in the past, where it describes nothing.

Figure (svg): Two panels. Left: the population line drawn against the calendar year, whose vertical intercept at year zero is about minus 755 thousand people. Right: the same line against years since 2004, whose intercept is the given 6,200.

Same town, same rate, two coordinate systems. Shifting the input to 2004 turns the intercept into a number the problem actually gave you.

\[ t = \text{years since } 2004 \quad\Longrightarrow\quad (0,\,6200) \text{ and } (5,\,8100) \]

Measuring years since 2004 makes the first data point the intercept, so the model can be written straight down with no extra work.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 234-235 — Example 2.26, the choice of variable

Look at the two panels side by side, because they make the same point twice. On the left, the input is the calendar year, and to reach the vertical intercept you have to run the line back more than two thousand years, where it predicts a population of about minus 755 thousand people. The model is not wrong there; it is simply far outside anything the data can speak to.

On the right, the input counts years since 2004. Now the intercept is 6,200, which is a number the problem gave you, and the model can be written down without computing anything beyond the slope.

The choice costs nothing and you make it, so make it well. Choose input zero at one of your data points, usually the earlier one, and you get that data point's output as the intercept for free. Then say in writing what the input means, because a reader who assumes a different starting year will read every one of your answers wrongly.

25. Estimate the 2013 population first

Estimation

A town grew from 6,200 people in 2004 to 8,100 in 2009, and the growth is linear.

Predict first

Before computing anything, estimate the population in 2013.

  • about 8,500
  • about 9,600
  • about 11,000
  • about 15,000

Correct: about 9,600

Why: The town gained 1,900 people in five years, so roughly 400 a year. From 2009 to 2013 is four more years, which is about 1,600 more people on top of 8,100: a little under 9,700. An estimate this rough is still enough to catch a slip of a factor of ten later on, which is the whole reason for making one.

Commit to an option before revealing. Estimating first is not a stalling tactic; it is how you find out later that you have made an arithmetic slip, because a wrong answer usually misses the estimate by a factor of ten rather than a few per cent.

The estimate here is a two-step mental calculation. Nineteen hundred people over five years is a bit under 400 a year. From 2009 to 2013 is four more years, so about 1,600 more people on top of 8,100, which lands just under 9,700.

That estimate is close enough to catch nearly any mistake you could make in the next two slides: using the wrong number of years, dividing the slope upside down, or forgetting that 2013 is nine years after 2004 rather than four. If your exact answer is not near your estimate, one of the two is wrong and you should find out which before moving on.

26. Example 2.26(a): the town's population in 2013

Worked example

A town's population was 6,200 in 2004 and 8,100 in 2009. Assuming the trend continues, predict the population in 2013.

Define the input so that zero is convenient

Why: Years since 2004, so the first fact sits at the intercept.

\[ t = \text{years since } 2004, \quad P(t) = \text{population} \]

Convert the two facts into coordinate pairs

Why: 2004 becomes 0 and 2009 becomes 5.

\[ (0,\,6200) \quad \text{and} \quad (5,\,8100) \]

Compute the slope

Why: Change in output over change in input.

\[ m = \frac{8100 - 6200}{5 - 0} = \frac{1900}{5} = 380 \]

Attach the units to the slope

Why: People divided by years.

\[ m = 380 \text{ people per year} \]

Write the model

Why: The first pair is the vertical intercept, so no further work is needed.

\[ P(t) = 380t + 6200 \]

Evaluate at the year asked for

Why: 2013 is nine years after 2004.

\[ P(9) = 380(9) + 6200 = 9620 \]

Figure (svg): The population line through the points 6,200 at t equals 0 and 8,100 at t equals 5, with a dashed step of 5 years across and 1,900 people up between them, and the predicted point 9,620 at t equals 9.

The dashed step is the slope: 1,900 people up for 5 years across, which is 380 people per year.

Check the model against the data it was built from

Why: Putting the second data point back in must return the second population.

\[ P(5) = 380(5) + 6200 = 8100 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 234-235 — Example 2.26a

Work through this one with the book closed if you can, because it is the archetype of the two-data-point problem and almost every exercise in the section is a version of it.

Two steps are worth slowing down on. The second step converts dates into inputs, and it is where an error quietly enters: 2009 becomes 5, not 2009 and not 9, because the input counts years after 2004. The sixth step converts back the other way, and 2013 is nine years after 2004, not four. Getting one of these right and the other wrong is a very common way to lose the whole question.

The check is the one you should copy into every problem of this type. A model built from two data points must return both of them. Putting the second point back in and getting 8,100 exactly tells you the slope and the intercept are consistent with the data you were given, which is the only part of the model that can be verified at all.

27. Example 2.26(b): the year the population reaches 15,000

Worked example

Using the same model, identify the year in which the population will reach 15,000.

Set the output to the target

Why: This is a solve, not an evaluation: the output is given and the input is wanted.

\[ 15000 = 380t + 6200 \]

Subtract the initial value

Why: Take 6,200 from both sides.

\[ 8800 = 380t \]

Divide by the rate

Why: Both sides by 380.

\[ t = \frac{8800}{380} \approx 23.158 \]

Convert the input back into a calendar year

Why: The input counts years after 2004.

\[ 2004 + 23.158 \approx 2027 \]

Figure (svg): The population line extended to t equals 27, with a dashed horizontal line at 15,000 people and a dashed vertical line at t equals 23.16 where the two meet.

Solving for the input is reading the graph sideways: come in at the height you want and drop down to the axis.

Check by putting the answer back into the model

Why: The model at 23.158 years must return the target population.

\[ P(23.158) = 380(23.158) + 6200 \approx 15000 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 235 — Example 2.26b

Same model, opposite direction. Part (a) gave you an input and asked for an output, so it was a substitution. Here you are given the output and asked for the input, so there is an equation to solve. Recognising which of the two you are facing is worth doing consciously until it becomes automatic.

The answer comes out as 23.158 years, and it should bother you slightly that a population reaches a round number partway through a year. It does not matter here, because the model was only ever an approximation, and the book rounds to say the population passes 15,000 somewhere around 2027. Say around, not in.

On the figure, notice how the answer was read: come in horizontally at the height you want, meet the line, and drop to the axis. That is what solving for the input looks like geometrically, and sketching it is a good way to check that your answer is on the right side of the data.

28. Trap: putting the calendar year into the model

Trap

The trap

The model counts years since 2004. Asked for 2013, a tempting first move:

\[ P(2013) = 380(2013) + 6200 \]

\[ = 771{,}140 \]

Wrong by a factor of eighty.

The fix

The input is not the date. It is the number of years after 2004, so the date has to be converted before it is substituted.

\[ t = 2013 - 2004 = 9 \]

\[ P(9) = 9620 \]

The same conversion runs the other way at the end of a solve: an answer of 23.158 is a count of years, and the calendar year is 2004 plus that.

This is the price of the clever choice made two slides ago. Shifting the input to years since 2004 made the model easy to write, and it also created a permanent obligation: every date has to be converted before it goes in, and every answer has to be converted back.

The wrong line is not a silly mistake; it is what happens when you stop thinking about what the input means and start treating it as a slot to put numbers in. The output, more than three quarters of a million people in a town of a few thousand, is eighty times too big, which is why the habit of glancing at an answer for plausibility is worth having.

Two defences. Write down in words what the input measures, right beside the model, so the conversion is staring at you. And when you report a final answer, convert it back and say it as a date: 2027, not 23.158.

29. Find the error: the slope upside down

Error analysis

A solution to Example 2.26 begins like this. The arithmetic is correct and the answer is still wrong.

Annotate

On: \( m = \frac{2009 - 2004}{8100 - 6200} = \frac{5}{1900} \)

  • Input change is on top and output change is underneath. The slope of a model is always output change over input change, so this fraction is the reciprocal of the slope.
  • This quantity is measured in years per person, which would answer 'how long until one more resident', not 'how many residents per year'. Checking units catches the error before the answer does.
  • Five over 1,900 is about 0.0026. A town adding three thousandths of a person a year should look wrong immediately.
  • Turn it over: 1,900 people over 5 years, which is 380 people per year. Then the model is 380t plus 6,200.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 235 — the slope formula in Example 2.26

Look at the fraction before revealing the annotations and try to name what is wrong with it. The arithmetic is flawless, which is exactly why this error survives all the way to a final answer.

Three separate checks would have caught it. The definition puts the change in output on top. The units come out as years per person, which does not answer a question about population growth. And the size, about three thousandths, is absurd for a town that grew by nineteen hundred people.

Of the three, the units check is the one to build into your routine, because it works even when you have no feeling for the numbers. Say the slope out loud with its units attached before you use it. If it sounds like an answer to a different question, you have the fraction upside down.

30. Try It 2.15: build the city's model

Faded example

A city's population has been growing linearly. In 2008 it was 28,200, and by 2012 it was 36,800. Let the input be years since 2008.

Fill in the blanks

P(t) = 2150t + 28200

Why: The population rose by 36,800 minus 28,200, which is 8,600 people, over 4 years, so the rate is 8,600 divided by 4, or 2,150 people per year. Because the input counts years since 2008, the 2008 population is the value at input zero and therefore the intercept, 28,200.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 235 — Try It 2.15

Fill in both blanks before checking, and do the arithmetic rather than eyeballing it. The population rose by 8,600 over four years, so the rate is 2,150 people per year.

The intercept needs no arithmetic at all, and that is the point of the instruction to let the input be years since 2008. The 2008 population is the output at input zero, so it is the intercept, exactly as it was in Example 2.26.

Before moving on, say what the two numbers mean in the city's own terms: the city gains about 2,150 people a year, and it had 28,200 people at the start of the period being modelled. If you can say that sentence, you understand the model; if you can only write the formula, you do not yet.

31. Try It 2.15: the 2014 prediction and the year of 54,000

Worked example

With the model just built, answer both parts of the Try It.

Recall the model and the meaning of its input

Why: Years since 2008, rate 2,150 people per year.

\[ P(t) = 2150t + 28200 \]

Predict 2014 by evaluating

Why: 2014 is six years after 2008.

\[ P(6) = 2150(6) + 28200 = 41100 \]

For the second part, set the output to the target

Why: Now the population is given and the year is wanted.

\[ 54000 = 2150t + 28200 \]

Subtract the initial value

Why: Take 28,200 from both sides.

\[ 25800 = 2150t \]

Divide by the rate

Why: It comes out to a whole number of years.

\[ t = \frac{25800}{2150} = 12 \]

Convert the input to a calendar year

Why: Twelve years after 2008.

\[ 2008 + 12 = 2020 \]

Figure (svg): The city population line through 28,200 at t equals 0 and 36,800 at t equals 4, with the 2014 prediction 41,100 at t equals 6 and the target 54,000 reached exactly at t equals 12.

Both data points are on the line by construction; the two answers are read off it, one by going up and one by coming across.

Check both answers in the model

Why: Each must return the number it came from.

\[ P(6) = 41100, \qquad P(12) = 2150(12) + 28200 = 54000 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 235 — Try It 2.15, both parts

Both parts use the model from the previous slide, and they are the two directions again: one substitution and one equation. Try them yourself before reading the steps.

The second part comes out to exactly 12 years, which is worth a comment. Whole-number answers in these problems are usually a sign that the exercise was designed backwards from the answer, not a sign that you are especially right. Check it rather than trusting it.

The check does exactly that, and it does it for both answers at once. Each computed value is put back into the model and must return the number it came from. Note also that the target of 54,000 lies beyond both data points, so this prediction is an extrapolation, a distinction you will make sharply in the next part of this lesson.

32. A data plan from two monthly bills

Worked example

Exercise 200. A phone company charges a flat monthly fee of 10 dollars plus an amount per megabyte of data. A customer who uses 20 megabytes pays 11.20 dollars; one who uses 130 megabytes pays 17.80. Find the model and the cost of 250 megabytes.

Convert the two bills into coordinate pairs

Why: Megabytes are the input, dollars the output.

\[ (20,\,11.20) \quad \text{and} \quad (130,\,17.80) \]

Compute the rate

Why: Change in dollars over change in megabytes.

\[ m = \frac{17.80 - 11.20}{130 - 20} = \frac{6.60}{110} = 0.06 \]

Attach the units

Why: Dollars per megabyte, which is six cents.

\[ m = 0.06 \text{ dollars per megabyte} \]

Use the flat fee as the intercept

Why: The problem gave the value at zero megabytes.

\[ C(x) = 0.06x + 10 \]

Answer the question asked

Why: Evaluate at 250 megabytes.

\[ C(250) = 0.06(250) + 10 = 25 \]

Figure (svg): The line C of x equals 0.06x plus 10 with the two known bills marked at 20 megabytes costing 11.20 and 130 megabytes costing 17.80, and the predicted 250 megabytes costing 25.00.

Two bills are two points, and two points fix a line. The flat fee was given, so it only has to be confirmed.

Check both given bills

Why: A model built from two points must reproduce both of them.

\[ C(20) = 11.20, \qquad C(130) = 17.80 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 244 — Exercise 200

This is the same structure as the population problems dressed in different clothes, and recognising that is most of the work. Two bills are two input-output pairs, and megabytes are the input because the bill follows from the usage.

There is one extra gift here: the problem states the flat monthly fee, which is the cost of using no data at all, and therefore the value at input zero. So the intercept does not have to be computed from a point, and the two bills are needed only for the rate.

That gift is also a trap, because it means the model is over-determined: you have three facts for two unknowns. The check exploits it. A correct model must reproduce both bills exactly, and if it reproduces only one of them, either the rate or the fee has been misread.

33. Put the modelling steps in order

Ranking

Put in order

Order the steps of building a linear model from two input-output pairs.

  1. Identify the input and output quantities, with units
  2. Convert the two facts into coordinate pairs
  3. Find the slope from the change in output over the change in input
  4. Write the linear model
  5. Evaluate it, or solve it, to answer the question posed
  6. Check the answer against the domain the situation allows

Why: Naming the quantities comes first because it fixes what a coordinate pair even means, and in particular where input zero sits. The slope cannot be computed before the pairs exist, and the model cannot be written before the slope. The last step is the one most often skipped, and it is the one that catches answers such as a negative number of months.

Drag the six steps into order before checking. Only two of the orderings are really in doubt, and both are worth thinking about.

The first is whether naming the quantities comes before converting the data into pairs. It must, because until you have said that the input counts years since 2008, the pair (0, 28200) does not mean anything at all.

The second is where the domain check goes. It is last, after you have an answer, because its job is to look at the answer and ask whether the model was entitled to produce it. Doing it first would be pointless; skipping it is how a model reports a baby at minus fifteen months.

34. Read the rate and the initial value out of each sentence

Comparison

Comparison matrix

the situationrateinitial value
saves 3,500, spends 400 a week-4003500
1,000 loan, pays 250 a month−2501000
25,000 fixed cost, 0.25 each0.2525000
7.5 pounds at birth, half a pound a month0.57.5
16 billion cubic feet, 2.1 used a year−2.116

Fill in each blank before checking. Every row is a situation from this lesson, so if one is unfamiliar, go back and find it rather than guessing.

The blanks in the first column are all rates, and two of them are negative. Check the sign against the verb before you type, and remember that a rate can be a fraction of a unit: 0.25 dollars per doughnut and half a pound per month are perfectly good rates.

The blanks in the last column are all initial values, and every one of them is the output before the input starts moving: the money saved, the loan before any payment, the reserve in the starting year. If you can read a table like this quickly, the modelling half of this lesson is done and only the domain questions are left.

35. The domain the situation allows

Section

Part 4

36. The formula accepts inputs the situation does not

Concept

When you model a real scenario, there is almost always a limited stretch of input over which the model is valid, because almost no trend continues indefinitely.

\[ 0 \le t \le 8.75 \]

For Emily, negative weeks would describe time before she had saved the money, and weeks beyond 8.75 would describe spending money she does not have. The formula computes happily at both. The situation does not.

reasonable domain — The set of input values for which the model actually describes the situation, as opposed to the set for which the formula is defined.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 233 — the reasonable domain for Emily's model

This is the idea that separates modelling from algebra, and it is the one most likely to appear on an exam as a question you cannot answer by computing anything.

The formula for Emily's money is defined for every real number, including minus three weeks and a hundred weeks. The situation is not. Before the trip she had not saved the money yet, and after 8.75 weeks she has none left, so unless she is willing to borrow, the model stops there.

Get into the habit of writing the reasonable domain on the same line as the model, every time. It costs you five seconds, it is often worth marks on its own, and it gives you something to check your final answers against. An answer outside the domain is not a small problem; it is a sign that either the model or the question has been misunderstood.

37. Two domains on one picture

Picture it

Figure (svg): Two number lines for the same variable. The upper one runs off in both directions with arrowheads, marked every real number. The lower one is a solid highlighted segment from 0 to 12 with closed ends, marked the months the baby has lived.

The two lines are not the same set. Every modelling answer has to be checked against the lower one.

The upper line is what the algebra will accept. The lower one is what the situation permits. Every answer you produce has to be checked against the lower line, and an answer that falls off it is not an answer at all.

The two lines in the picture describe the same variable and they are not the same set. That is the whole slide. The upper line is the formula's domain, running off in both directions with no end; the lower one is the situation's, a closed segment with two ends you can name.

The red dot on the upper line is where the weight formula's horizontal intercept sits, at minus fifteen months. There is nothing wrong with the formula for producing it. It is simply a point the situation never reaches, which is why it has no interpretation.

When you finish a problem, put your answer on the lower line and see whether it lands inside the segment. If it does, report it. If it does not, the right answer is a sentence: the model reaches that value outside the range it describes, so it cannot be used to answer the question.

38. The newborn's weight: a reasonable domain and range

Worked example

Exercises 169 to 173. A newborn weighs 7.5 pounds and gains half a pound a month for its first year. Model the weight, state a reasonable domain and range, and find when the baby weighed 10.4 pounds.

Write the model

Why: Rate one half pound per month, initial value 7.5 pounds.

\[ W(t) = 0.5t + 7.5 \]

State the domain the situation allows

Why: The description covers the first year only.

\[ 0 \le t \le 12 \]

Find the range from the endpoints

Why: A line is increasing here, so the two ends give the two extremes.

\[ W(0) = 7.5, \quad W(12) = 13.5 \]

Write the range

Why: Everything between those two weights.

\[ 7.5 \le W \le 13.5 \]

Set the output to the weight asked about

Why: Now the weight is given and the month is wanted.

\[ 10.4 = 0.5t + 7.5 \]

Solve for the month

Why: Subtract, then divide by one half.

\[ 2.9 = 0.5t \;\Longrightarrow\; t = 5.8 \]

Figure (svg): The line W equals 7.5 plus 0.5t drawn from t equals minus 18 to t equals 16 months, with the stretch from 0 to 12 shaded, the point 10.4 pounds at 5.8 months marked, and the horizontal intercept at minus 15 months marked in red.

The formula is happy at minus 15 months. The baby is not. Only the shaded stretch is the model.

Check that the answer lies inside the domain

Why: 5.8 months is within the first year, so the model was entitled to answer; had it come out at 20 months the answer would have had to be thrown away.

\[ 0 \le 5.8 \le 12 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 242 — Exercises 169 to 173

This exercise exists to make you state a domain and a range, so do not rush past those two steps to get to the arithmetic. The domain is a set of inputs, measured in months, and the description gives it to you: the first year, so zero to twelve.

The range is a set of outputs, measured in pounds, and for a linear function on a closed interval you get it from the two endpoints, because a line is increasing or decreasing throughout and cannot turn around. Here the weight runs from 7.5 pounds at birth to 13.5 pounds at one year.

The last part of the question is another solve, and the answer, 5.8 months, has to be checked against the domain before being reported. It is inside, so the model may answer. Look at the figure and notice how much of the line lies outside the shaded stretch: that is how much of the formula has nothing to do with this baby.

39. The weight model's horizontal intercept

Anomaly

\[ W(t) = 0.5t + 7.5 \quad\Longrightarrow\quad W(t) = 0 \text{ at } t = -15 \]

Predict first

The weight model has a horizontal intercept at minus 15 months. What is the right thing to say about it?

  • The baby weighed nothing fifteen months before birth
  • The model does not apply there, so the intercept has no meaning
  • The model is wrong and must be rebuilt
  • The weight was negative before birth

Correct: The model does not apply there, so the intercept has no meaning

Why: The intercept is a genuine feature of the formula and a meaningless feature of the situation, because minus 15 months is outside the reasonable domain. The model is not wrong: it was only ever claimed to describe the first year. This is why a reasonable domain is stated. Compare Emily's model, where the horizontal intercept sits inside the domain and answers the actual question.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 242 — Exercise 171, interpreting the intercepts

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 233 — where a model stops being valid

Commit to an option before revealing. This question sorts people who have understood reasonable domains from people who have only learned the phrase.

The intercept is real: the formula genuinely crosses zero at minus fifteen months. What it is not is meaningful, because minus fifteen months is outside the domain the situation allows, and the model never claimed to describe anything there. The correct thing to write on an exam is that the horizontal intercept lies outside the reasonable domain and therefore has no interpretation in this context.

Contrast this with Emily's model, where the horizontal intercept sat at 8.75 weeks, inside the domain, and was the answer to the question. Whether an intercept means something depends entirely on whether it lands inside the domain, which is another reason to write the domain down.

40. The federal helium reserve: model and end date

Worked example

Exercise 203. The Federal Helium Reserve held about 16 billion cubic feet of helium in 2010, and is being depleted by about 2.1 billion cubic feet each year. Give the model, the 2015 reserve, and the year of depletion.

Name the input and write the model

Why: Years since 2010, billions of cubic feet remaining.

\[ R(t) = 16 - 2.1t \]

Evaluate at 2015

Why: Five years after 2010.

\[ R(5) = 16 - 2.1(5) = 5.5 \]

Set the reserve to zero for the second question

Why: Depleted means nothing remaining.

\[ 0 = 16 - 2.1t \]

Solve for the year count

Why: Divide 16 by the yearly loss.

\[ t = \frac{16}{2.1} \approx 7.62 \]

Convert to a calendar year

Why: About seven and a half years after 2010.

\[ 2010 + 7.62 \approx 2017.6 \]

Figure (svg): The line R equals 16 minus 2.1t falling from 16 billion cubic feet at t equals 0 to zero at about t equals 7.62, with the 2015 value 5.5 marked.

A constant drawdown makes a straight line, and a straight line that falls has exactly one crossing: the year the tap runs dry.

Check by bracketing the crossing

Why: The reserve is still positive at seven years and negative at eight, so depletion falls between them, during 2017.

\[ R(7) = 1.3 > 0, \qquad R(8) = -0.8 < 0 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 244 — Exercise 203

This problem is worth doing because it is exactly the shape of a great many real questions: a stock, a constant drawdown, and the two things people want to know, namely how much is left at a given time and when it runs out.

The two parts are again a substitution and a solve. The 2015 reserve is an evaluation at five years. The depletion date sets the output to zero and solves for the input, which is the horizontal intercept of the model.

Look at the final check, because it uses a technique worth having. Rather than trusting the division, it brackets the answer: the reserve is still positive after seven years and negative after eight, so the crossing is somewhere between them, during 2017. Bracketing catches sign errors and arithmetic slips at once, and it also gives you a way to answer in whole years, which is usually what the question wants.

41. Trap: answering outside the stated domain

Trap

The trap

A town of 75,000 grows at 2,500 a year for 5 years. When does it reach 100,000?

\[ 100000 = 2500t + 75000 \]

\[ t = 10 \text{ years} \]

Wrong, or at least not an answer: the model was only claimed for five years.

The fix

The algebra is fine. The model is not on the hook at ten years, because the description ran out at five. The honest answer names the gap: the model reaches 100,000 at ten years, which is outside the five-year window it was built for, so it is a guess and not a prediction.

\[ 0 \le t \le 5 \]

\[ P(5) = 87500 \]

Inside the window the town only gets to 87,500. Anything past that is extrapolation.

Read the wrong side carefully, because there is nothing wrong with the algebra. The model does reach 100,000 at ten years. The problem is that the description said the town grows at that rate for five years, and the model was only ever claimed for those five.

What should you write instead? Not nothing, and not the bare number. Write that the model reaches 100,000 at ten years, which is outside the five-year window the model describes, so it is an extrapolation rather than a prediction. That answer shows you can do the algebra and that you know what it is worth.

This particular exercise is set deliberately, and so is its neighbour that asks for the output at twelve years. Both are testing whether you notice the phrase for five years. When a problem tells you how long a trend lasts, it is telling you the domain, and it expects you to use it.

42. Between the data, and beyond it

Concept

Figure (svg): A line through two data points at x equals 3 and x equals 9, with the band between them shaded and labelled interpolation and the regions to the left and right labelled extrapolation.

Inside the band the line is pinned down by data on both sides. Outside it, nothing holds the line to reality.

interpolation — Using a model at an input that lies between the data points it was built from.

extrapolation — Using a model at an input outside that range, where no data constrains it.

Both are legitimate and both are done constantly, but they carry different risks. Between two data points the line is pinned at both ends. Beyond them nothing holds it down, and the further out you go the more the answer depends on the assumption that the trend continues.

model breakdown — The point at which a model stops describing the situation, so that its output becomes unreliable or absurd.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 233 — almost no trend continues indefinitely

Both words describe using the model at an input it was not built at, so the difference between them is about where that input sits. Inside the data, the line is anchored at both ends and small errors cannot grow far. Outside, nothing holds it, and the error grows with the distance.

Look at the figure and imagine wiggling one of the two data points slightly. Inside the shaded band the line barely moves. Far out to the right, the same wiggle swings the line by a large amount, which is a concrete reason to trust extrapolation less.

Model breakdown is the third term and the most important. It names the fact that a trend eventually stops: the town stops having room, the baby stops growing half a pound a month, the reserve runs out. When you extrapolate, the real question is never whether the arithmetic is right but whether the trend has broken down before you get there.

43. Indiana and Alabama home values: an absurd prediction

Worked example

Exercise 196. Median home values, adjusted for inflation, were 37,700 in Indiana and 27,100 in Alabama in 1950, and 94,300 and 85,100 in 2000. Assuming both change linearly, in what year are they equal?

Define one input for both models

Why: Years since 1950, so both start at the intercept.

\[ t = \text{years since } 1950 \]

Compute Indiana's rate

Why: Change in value over fifty years.

\[ m_I = \frac{94300 - 37700}{50} = 1132 \]

Compute Alabama's rate

Why: The same calculation for the second state.

\[ m_A = \frac{85100 - 27100}{50} = 1160 \]

Write both models

Why: Each has its own 1950 value as its intercept.

\[ I(t) = 1132t + 37700, \quad A(t) = 1160t + 27100 \]

Set them equal

Why: Equal home values means equal outputs.

\[ 1132t + 37700 = 1160t + 27100 \]

Collect the terms

Why: Take the smaller rate and the smaller intercept across.

\[ 10600 = 28t \]

Solve, then convert to a year

Why: Alabama closes a 10,600 dollar gap at 28 dollars a year.

\[ t = \frac{10600}{28} \approx 378.6 \;\Longrightarrow\; 1950 + 378.6 \approx 2329 \]

Figure (svg): Two panels. Left: the Indiana and Alabama home-value lines over the fifty years of data, close together and slowly converging. Right: the gap between them, 10,600 dollars shrinking by 28 dollars a year, hitting zero only after 378 years.

The data cover the left panel only. Everything to the right of t equal to 50 is the model talking to itself.

Check the arithmetic, then the sense of the answer

Why: Both models do give the same value there, about 466,000 dollars, so the algebra is right; but the answer extrapolates fifty years of data across four centuries, which is model breakdown, not a prediction.

\[ I(378.6) \approx 466{,}200 \approx A(378.6) \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 243 — Exercise 196, which warns that the answer might be absurd

The exercise itself warns you that the answer might be absurd, which is a rare and useful hint. Do the algebra anyway, then look hard at what comes out.

The two states are separated by 10,600 dollars in 1950, and Alabama closes that gap at only 28 dollars a year, because the two growth rates are nearly the same: 1,132 against 1,160. A small gap divided by a much smaller closing speed gives an enormous number of years, which is why the answer lands in the year 2329.

Two lessons come out of this. The first is arithmetical: when you subtract two nearly equal rates, tiny changes in either one move the answer enormously, so the result is fragile even on its own terms. The second is about modelling: fifty years of data have been run out over four centuries. Look at the right panel of the figure and see how little of it is supported by anything. That is model breakdown, and the honest answer says so.

44. Interpolation or extrapolation?

Discrimination

Sort into buckets

Each model was built from the two data points named. Sort the uses.

interpolation
town population in 2007, from 2004 and 2009; city population in 2010, from 2008 and 2012; Indiana home value in 1975, from 1950 and 2000
extrapolation
town population in 2013, from 2004 and 2009; Indiana home value in 2010, from 1950 and 2000; the year Indiana and Alabama values agree
in
2007 sits between 2004 and 2009; 2010 sits between 2008 and 2012; 1975 sits between 1950 and 2000. In each case data sit on both sides of the input, so the line is held down at both ends.
out
2013, 2010 and the year 2329 all lie beyond the last data point. They may still be reasonable, but each rests entirely on the assumption that the trend continues, and the last one rests on it for nearly four centuries.

Sort all six before checking, and use one test for each: does the input lie between the two data points the model was built from, or outside them?

The three that are interpolations are all comfortable. 2007 sits between 2004 and 2009, 2010 between 2008 and 2012, and 1975 between 1950 and 2000. In each case the line is held down on both sides of the answer.

The three extrapolations are not equally bad, and it is worth ranking them. Predicting 2013 from data ending in 2009 is four years out and probably fine. Predicting a home value in 2010 from data ending in 2000 is a decade out and starting to be a guess. Predicting the year 2329 is arithmetic with no claim on reality at all. Extrapolation is a matter of degree, not a yes or no.

45. The world's oil reserves

Real world

\[ R(t) = 1820 - 25t \]

Discussion prompt

Exercise 204 says the world's oil reserves in 2014 were about 1,820 billion barrels, falling by about 25 billion barrels a year. Give the reserves seven years later, the year of depletion, and one reason the second answer is far less trustworthy than the first.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 244 — Exercise 204

Write both answers before revealing, and then spend most of your time on the third part of the question, which is the one that is actually being taught here.

The arithmetic is the same shape as the helium problem: evaluate at seven years, then set the output to zero and solve. You should get 1,645 billion barrels and about 72.8 years.

The reason the second answer deserves much less trust is that it depends on the drawdown rate staying exactly constant for nearly three quarters of a century, over which new fields are found, consumption changes and other energy sources grow. Seven years out, the same assumption is fairly safe. Same model, same algebra, two very different levels of confidence, and the difference is entirely the distance from the data.

46. When a diagram is the model

Section

Part 5

47. Draw the situation before choosing letters

Concept

Some problems describe positions rather than amounts. There the picture comes first: sketch the figure, mark the quantities you know, and give a letter to each thing that changes. Geometry then supplies the relationship that words alone were hiding.

\[ a^2 + b^2 = c^2 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 236 — using a diagram to model a problem

Some problems in this section are not about amounts at all but about positions, and for those the diagram is not an illustration; it is the step that produces the relationship. Words can hide a right angle. A sketch cannot.

The second bullet is the one people skip. Saying where the origin is and which direction counts as positive is what makes a pair of coordinates mean something, and in the road problem it is what lets you write down two equations instead of staring at a map.

The last two bullets are the two facts from earlier chapters that these problems lean on. Any right triangle brings the Pythagorean theorem with it, and any two perpendicular lines have slopes that are negative reciprocals. Both of those appear in the next three worked examples.

48. Example 2.27: when do Anna and Emanuel lose radio contact?

Worked example

Anna and Emanuel start at the same intersection. Anna walks east at 4 miles per hour, Emanuel south at 3 miles per hour. Their two-way radio has a range of 2 miles. How long until they fall out of contact?

Name the input and the outputs

Why: Time drives everything else, so it is the input.

\[ t = \text{hours}, \quad A(t), E(t) = \text{miles walked} \]

Read the initial values

Why: Both start at the intersection, so both distances start at zero.

\[ A(0) = 0, \quad E(0) = 0 \]

Write each distance walked

Why: Constant speed means a constant rate of change.

\[ A(t) = 4t, \qquad E(t) = 3t \]

Figure (svg): A right triangle. From the starting intersection an arrow runs right, labelled Anna 4t miles east, and another runs down, labelled Emanuel 3t miles south. The slanted side joining their positions is labelled D of t equals 5t, and a small square marks the right angle at the intersection.

The two walkers make the legs; the radio has to reach along the slanted side. Three, four, five is doing the work here.

Name the quantity the question is really about

Why: The radio cares about the straight-line distance between them.

\[ D(t) = \text{miles between Anna and Emanuel} \]

Use the right angle at the intersection

Why: East and south are perpendicular, so the two paths are the legs of a right triangle.

\[ D(t)^2 = A(t)^2 + E(t)^2 = (4t)^2 + (3t)^2 \]

Simplify the right side

Why: Square each term and add.

\[ D(t)^2 = 16t^2 + 9t^2 = 25t^2 \]

Take the square root, keeping time positive

Why: Only positive t is considered, so the absolute value can be dropped.

\[ D(t) = 5t \]

Set the distance to the radio range

Why: Contact is lost at exactly 2 miles apart.

\[ 5t = 2 \;\Longrightarrow\; t = 0.4 \]

Convert to minutes

Why: Four tenths of an hour.

\[ 0.4 \text{ h} = 24 \text{ min} \]

Check the answer against the triangle

Why: Their positions at that moment must be exactly 2 miles apart.

\[ \sqrt{1.6^2 + 1.2^2} = \sqrt{2.56 + 1.44} = \sqrt{4} = 2 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 236-238 — Example 2.27

Start by asking what the question is really about. It sounds like it is about two walkers, but the radio only cares about one quantity: the straight-line distance between them. Turning the question into how long until they are two miles apart is the first real step, and the diagram is what makes that turn visible.

Both walkers start at the same intersection, so both distances start at zero and both models are simply speed times time. Because east and south are at right angles, those two distances are the legs of a right triangle, and the distance between the walkers is its hypotenuse.

The pleasant surprise is the square root: the sum of 16 t squared and 9 t squared is 25 t squared, whose square root is 5 t. The distance between them is itself a linear function of time, with slope 5 miles per hour, so the final question needs one small equation. Note the step where the absolute value is dropped: only positive times are being considered, which is a statement about the domain.

49. The distance between them is linear too

Picture it

Figure (svg): The line D equals 5t rising from the origin, crossing a dashed horizontal line at 2 miles when t equals 0.4 hours.

Two linear distances at right angles give a third linear distance, so the last step is one small equation.

This is the pay-off of the diagram. Each walker's distance from the intersection is linear in time, and because they are walking at right angles the distance between them turns out to be linear as well, with slope 5. So the final question is one small equation rather than anything harder.

The three, four, five triangle is not a coincidence of this problem: any two perpendicular constant speeds give a constant separation speed, namely the square root of the sum of their squares.

This figure is here to make one idea concrete. The graph of the distance between the walkers is a straight line through the origin with slope 5, so they separate at a steady 5 miles per hour even though neither of them is walking at that speed.

The answer is read off the picture the same way as before: come in at the radio range of 2 miles, meet the line, and drop to the axis at 0.4 hours. Converting to 24 minutes is worth doing, because a time in hours with a decimal point is not how anybody answers this question out loud.

The second paragraph on the slide is the general fact worth remembering. Two perpendicular constant speeds always combine into a constant separation speed, given by the square root of the sum of their squares. Here the three, four, five triangle makes it come out whole, which is why the book chose those speeds.

50. Example 2.28: how far is the junction from Westborough?

Worked example

A straight road runs from Westborough to Agritown, 30 miles east and 10 miles north. Partway along, a second road, perpendicular to the first, leads to Eastborough, which is 20 miles directly east of Westborough. How far is the junction from Westborough?

Put the origin where it helps most

Why: Westborough is the town the question measures from.

\[ W = (0,\,0), \; A = (30,\,10), \; E = (20,\,0) \]

Find the slope of the Agritown road

Why: Ten miles north for thirty miles east.

\[ m = \frac{10 - 0}{30 - 0} = \frac13 \]

Write that road's equation

Why: It passes through the origin, so there is no constant term.

\[ W(x) = \tfrac13 x \]

Use perpendicularity for the second road

Why: Perpendicular slopes are negative reciprocals.

\[ m_{\perp} = -3 \]

Find its constant from the Eastborough point

Why: Substitute the point into the slope-intercept form.

\[ 0 = -3(20) + b \;\Longrightarrow\; b = 60 \]

Write the second road's equation

Why: Both roads now have equations.

\[ E(x) = -3x + 60 \]

Set them equal to find the junction

Why: The junction is on both roads.

\[ \tfrac13 x = -3x + 60 \]

Clear the fraction

Why: Multiply every term by 3.

\[ x = -9x + 180 \]

Solve for the first coordinate

Why: Add 9x to both sides, then divide by 10.

\[ 10x = 180 \;\Longrightarrow\; x = 18 \]

Find the second coordinate

Why: Substitute back into the simpler road.

\[ y = \tfrac13(18) = 6 \]

Figure (svg): A map with Westborough at the origin, Agritown 30 miles east and 10 miles north, and Eastborough 20 miles due east. The straight road from Westborough to Agritown is drawn, and a shorter road perpendicular to it runs from Eastborough up to the junction at 18 miles east and 6 miles north.

Choosing Westborough as the origin is what makes the two road equations easy; the junction is just where they meet.

Check with the distance formula

Why: The question asked for a distance from Westborough, which is the origin.

\[ d = \sqrt{(18-0)^2 + (6-0)^2} = \sqrt{360} \approx 18.974 \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 238-239 — Example 2.28

This is the longest worked example in the lesson and every line of it is Section 2.2 material, so read it as a chain of small familiar moves rather than as one big problem.

The first decision does the most work. Putting the origin at Westborough turns the three towns into coordinates, one of which is the origin itself, so the first road passes through the origin and its equation has no constant term. Choosing any other origin would give the same answer with more arithmetic.

Watch the two places where the geometry enters. The perpendicular road's slope is the negative reciprocal of one third, which is minus three; and the junction is where the two roads meet, which is where their outputs are equal. Then the distance formula finishes it. If the last step feels like it came from nowhere, remember that the distance formula is just the Pythagorean theorem applied to the horizontal and vertical gaps between two points.

51. Try It 2.16: the junction on the Timpson road

Worked example

A straight road runs from Timpson to Ashburn, 60 miles east and 12 miles north. A perpendicular road leaves it for Garrison, 22 miles directly east of Timpson. How far is the junction from Timpson?

Place the origin at Timpson

Why: The same construction as Example 2.28.

\[ T = (0,\,0), \; A = (60,\,12), \; G = (22,\,0) \]

Find the slope of the Ashburn road

Why: Twelve north for sixty east.

\[ m = \frac{12}{60} = \frac15 \]

Write its equation

Why: Through the origin again.

\[ R(x) = \tfrac15 x \]

Take the negative reciprocal

Why: The Garrison road is perpendicular.

\[ m_{\perp} = -5 \]

Find its constant from the Garrison point

Why: Substitute the point (22, 0).

\[ 0 = -5(22) + b \;\Longrightarrow\; b = 110 \]

Set the two equations equal

Why: The junction lies on both roads.

\[ \tfrac15 x = -5x + 110 \]

Clear the fraction and collect

Why: Multiply through by 5, then gather the x terms.

\[ x = -25x + 550 \;\Longrightarrow\; 26x = 550 \]

Solve for both coordinates

Why: Divide, then substitute back.

\[ x = \tfrac{275}{13} \approx 21.15, \quad y = \tfrac15 x = \tfrac{55}{13} \approx 4.23 \]

Figure (svg): A map with Timpson at the origin, Ashburn 60 miles east and 12 miles north, and Garrison 22 miles due east. The Ashburn road is drawn from the origin, and a short perpendicular road runs from Garrison up to the junction at about 21.15 miles east and 4.23 miles north.

Identical construction, different numbers: slope one fifth, perpendicular slope minus five, then one intersection.

Check with the distance formula

Why: The distance from the origin to the junction, to two decimal places.

\[ d = \sqrt{21.15^2 + 4.23^2} \approx 21.57 \text{ miles} \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 239 — Try It 2.16

Do this one yourself with the previous example beside you, because it is the same construction with different numbers and nothing new in it. That makes it the best possible check on whether you followed Example 2.28 or only watched it.

The numbers are less friendly here. The slope is one fifth, the perpendicular slope is minus five, and the junction lands at 275 over 13, which is about 21.15. Keep the fraction until the final step and round only once, at the end, or the distance will come out slightly wrong.

Round the final answer to 21.57 miles and notice that it is just under 22, the distance to Garrison. That is a good sanity check: the junction is a little closer to Timpson than Garrison is, which is exactly what the map shows.

52. The slope of the perpendicular road

Prediction

\[ R(x) = \tfrac15 x \]

Predict first

A road has the equation above. A second road crosses it at a right angle. What is the second road's slope?

  • negative five
  • one fifth
  • negative one fifth
  • five

Correct: negative five

Why: Perpendicular slopes multiply to negative one, so the second slope is the negative reciprocal of one fifth. Turn the fraction over to get five and change the sign to get negative five. A road that climbs gently as it runs east is crossed by one that drops steeply, which is exactly what the map shows.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 238 — the perpendicular road in Example 2.28

Commit to one option before revealing. There are only four and two of them have the wrong sign, so the real question is whether you remember that perpendicular slopes are negative reciprocals rather than just negatives.

The reciprocal of one fifth is five, and the sign flips, giving minus five. A quick way to confirm it: multiply the two slopes together. One fifth times minus five is minus one, which is the test for perpendicularity from Section 2.2.

Now picture it on the map. A slope of one fifth is a road that climbs gently as it runs east; a slope of minus five is a road that drops steeply. Steep and gentle is what perpendicular looks like when one of the two lines is nearly flat, and seeing that on the figure is a good way to remember which of the two numbers is which.

53. Two models, one decision

Section

Part 6

54. Setting two models equal

Concept

When a situation offers a choice between two plans, each plan gets its own model with the same input. Setting the two outputs equal finds the input at which the plans agree, and that input splits every other input into two ranges: one where the first plan wins and one where the second does.

Figure (svg): Three small panels of two lines each: crossing at one point, parallel and never meeting, and lying exactly on top of each other.

Two linear models can agree once, never, or always. Almost every word problem is the first case, but the other two are worth recognising.

\[ K(d) = M(d) \quad\Longrightarrow\quad d = \text{the break-even input} \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 239-240 — building systems of linear models, Figure 2.42

The new idea here is small and the consequences are large. Two competing plans give two models with the same input, and setting the outputs equal finds the input where they agree.

What makes it useful is what happens on either side of that input. Because both models are straight lines, one of them is above the other everywhere to the left of the crossing and below it everywhere to the right. So a single crossing splits every possible input into two ranges, and one calculation answers the question for all of them.

The three panels show the three ways two lines can sit. Almost every word problem gives you the first: two different rates, one crossing. Parallel lines mean the plans differ by a constant amount for ever, so one is always better, and identical lines mean the two plans are the same plan written twice.

55. Fee against rate: who wins in the end

Intuition

In almost every comparison of this kind, one plan has the higher fee and the other has the higher rate. Subtracting the two models shows exactly how the contest goes.

\[ K(d) - M(d) = (0.59d + 20) - (0.63d + 16) = 4 - 0.04d \]

The difference is itself a linear function. It starts at the gap between the two fees, 4 dollars, and it closes at the gap between the two rates, 4 cents per mile. Divide one by the other and you have the break-even input without ever setting the models equal.

So the plan with the lower rate always wins eventually, and the only question is how much input it takes. A large fee gap and a small rate gap means a long wait, which is exactly the situation in the Indiana and Alabama problem.

Subtracting one model from the other is a move worth having, because it turns a comparison of two lines into a question about one. The difference of two linear functions is itself linear, and its intercept and slope both mean something.

Its intercept is the gap between the two fees, which says who is ahead before anything has happened. Its slope is the gap between the two rates, which says how fast that lead is being eaten away. The break-even input is just the first divided by the second, and here that is 4 dollars divided by 4 cents a mile, giving 100 miles.

The general conclusion is worth remembering for the rest of your life, not just this exam: the option with the lower per-unit rate always wins eventually, and the only question is how much use it takes to get there. Phone plans, memberships, printers and electric cars are all this problem.

It also explains the Indiana and Alabama answer. There the fee gap was 10,600 dollars and the rate gap only 28 dollars a year, so the wait was 378 years. A big gap closing at a slow rate takes an absurdly long time.

56. Reading the comparison as an inequality

Notation

Annotate

On: \( K(d) < M(d) \;\iff\; 0.59d + 20 < 0.63d + 16 \)

  • Cheaper means a smaller cost, so 'when is Keep on Trucking the better choice' is the inequality with Keep's cost on the smaller side.
  • Both sides use the same d. Comparing two models only makes sense when they are fed the same input in the same units.
  • The answer to an inequality is a range of inputs, not a single number. The equation gives the one input where the two agree, and the range starts there.
  • Keep starts higher but climbs more slowly, so it must be the expensive one at first and the cheap one later. Testing one convenient value settles which side is which.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 240-241 — Example 2.29

The annotations are about turning a sentence into a symbol, which is where these questions are won and lost. Better choice means cheaper, cheaper means a smaller output, so the company you are asking about goes on the smaller side of the inequality.

The third annotation names the shape of the answer, and it is worth saying out loud: the answer to a question like this is a range of inputs, not a single number. The number you compute is the boundary of that range, and it is the one input at which neither plan is better.

The fourth annotation is the practical trick. Rather than doing the algebra of an inequality, solve the equation, then test one convenient input on either side of the answer. That is what the worked example does next, and it is faster and much harder to get backwards.

57. Example 2.29: which truck rental is cheaper?

Worked example

Keep on Trucking charges 20 dollars up front, then 59 cents a mile. Move It Your Way charges 16 dollars up front, then 63 cents a mile. When is Keep on Trucking the better choice?

Name the input and the two outputs

Why: One input, two costs, so two functions.

\[ d = \text{miles driven}, \quad K(d), M(d) = \text{cost in dollars} \]

Write both models from the fees and the rates

Why: The up-front fee is each model's initial value.

\[ K(d) = 0.59d + 20, \qquad M(d) = 0.63d + 16 \]

Set them equal to find the break-even mileage

Why: The costs agree where the two lines cross.

\[ 0.59d + 20 = 0.63d + 16 \]

Collect the terms

Why: Subtract 0.59d and 16 from both sides.

\[ 4 = 0.04d \]

Solve

Why: Divide by four hundredths.

\[ d = 100 \text{ miles} \]

Decide which side of 100 belongs to which company

Why: Keep climbs more slowly, so it wins on the longer trips.

\[ K(150) = 108.50 < 110.50 = M(150) \]

Figure (svg): The two rental cost lines, Keep on Trucking starting at 20 dollars with the shallower slope and Move It Your Way starting at 16 with the steeper slope, crossing at 100 miles and 79 dollars, with the region beyond 100 miles shaded.

The company with the higher fee and the lower rate wins in the long run; the crossing says exactly how long the long run is.

Check the other side too

Why: Below the break-even the cheaper up-front fee should win, and it does.

\[ K(50) = 49.50 > 47.50 = M(50) \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 240-241 — Example 2.29

The setup is the same as every other model in this lesson, twice over: two up-front fees are two initial values, and two per-mile charges are two rates. The only new step is the one where the two models are set equal.

There is a wording slip in the printed solution here. The book says it is looking for when Move It Your Way will cost less, but the inequality it writes, and the question it answers, are about Keep on Trucking being cheaper. The inequality in the book is the correct one; the sentence beside it is not. This is a good reminder to trust the mathematics you can check over the words around it.

The two checks at the end are the real answer to the question. Keep on Trucking has the higher fee and the lower rate, so it must lose on short trips and win on long ones, and testing 50 miles and 150 miles confirms both halves. Whenever you finish a comparison problem, test one value on each side of the crossing; it takes ten seconds and it fixes the direction of the inequality for good.

58. Two pay options: how much must you sell?

Worked example

Exercise 206. Option A pays a base salary of 17,000 dollars a year plus 12 percent commission on sales. Option B pays 20,000 plus 5 percent. How much must you sell for option A to pay more?

Name the input and the two outputs

Why: Yearly sales drive both incomes.

\[ s = \text{yearly sales in dollars} \]

Turn each percentage into a rate

Why: Twelve percent is 0.12 dollars of income per dollar of sales.

\[ A(s) = 17000 + 0.12s, \quad B(s) = 20000 + 0.05s \]

Set the two incomes equal

Why: Find where the options pay the same.

\[ 17000 + 0.12s = 20000 + 0.05s \]

Collect the terms

Why: Subtract 0.05s and 17,000 from both sides.

\[ 0.07s = 3000 \]

Solve

Why: Divide by seven hundredths.

\[ s = \frac{3000}{0.07} \approx 42857.14 \]

State the answer as a range

Why: Option A has the steeper rate, so it pulls ahead above the crossing.

\[ A(s) > B(s) \iff s > 42857.14 \]

Figure (svg): Two income lines against yearly sales in thousands of dollars: option A starting at 17 thousand with the steeper slope and option B starting at 20 thousand with the shallower slope, crossing at about 42.9 thousand in sales.

Option B starts 3,000 ahead; option A gains 70 dollars on it for every thousand sold, so it needs about 42,900 of sales to catch up.

Check one value on each side

Why: Just above the crossing A must win, and just below it B must.

\[ A(50000) = 23000 > 22500 = B(50000) \]

\[ A(40000) = 21800 < 22000 = B(40000) \]

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 245 — Exercise 206

Commission problems look different from rental problems and are not. A base salary is an initial value and a commission rate is a rate of change, the only difference being that the rate is written as a percentage and has to be converted to a decimal before it can be used.

Twelve percent means 0.12 dollars of income for every dollar of sales, so the slope is 0.12, and it is measured in dollars per dollar, which is why it looks like a pure number. Option B starts 3,000 dollars ahead and option A gains 0.07 dollars on it per dollar sold, so A needs 3,000 divided by 0.07 in sales, about 42,857 dollars, to catch up.

The answer is a range, not a number: option A pays more for sales above that figure. And the check tests both sides, which matters because it is easy to state the range backwards. If you sell very little, the bigger base salary wins; if you sell a lot, the bigger commission does.

59. Slide the commission rate

Tweak it

Parameter explorer

The curve is option A's income minus option B's, so the plans agree where it crosses the horizontal axis. Slide the commission rate of option A and watch the crossing move. At what commission rate does option A never catch up?

\[ A(s) - B(s) = -3000 + \left(\frac{{c}}{100} - 0.05\right)s \]

  • c — from 2 to 16: option A commission, percent

Move the slider slowly from 12 downward and watch the crossing point slide to the right. The curve is the difference between the two incomes, so where it crosses the horizontal axis the two options pay the same, and to the right of that point option A is ahead.

At a commission of exactly five percent something qualitative happens: the two options rise at the same rate, the difference becomes a flat line 3,000 dollars below zero, and there is no crossing at all. Option A can never catch up, no matter how much you sell, because it never gains on option B.

Below five percent the difference line slopes downward, and the harder you sell the worse option A is. That is the whole of this part of the lesson in one picture: the intercepts say who is ahead at the start, and the rates say who wins in the end.

60. Window washing: which company wins?

Prediction

One company charges 5 dollars per window. A second charges a base fee of 40 dollars plus 3 dollars per window.

Predict first

For how many windows is the second company the cheaper choice?

  • more than 20 windows
  • exactly 20 windows
  • fewer than 20 windows
  • more than 8 windows

Correct: more than 20 windows

Why: Setting the two costs equal gives five times the number of windows equal to forty plus three times it, so two times the number is forty and the number is twenty. The second company charges less per window but starts 40 dollars behind, so it only wins once there are enough windows to make up the base fee: above twenty. At exactly twenty both cost 100 dollars.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 244 — Exercise 205

Choose an option before revealing, and try to do it without writing anything: the numbers are small enough for mental arithmetic and the reasoning is the point.

The second company charges 2 dollars less per window but starts 40 dollars behind, so it needs 40 divided by 2, which is 20 windows, to make up the difference. Above twenty windows it is cheaper; below, the first company is; at exactly twenty both charge 100 dollars.

The tempting wrong answer is fewer than twenty, chosen by people who correctly compute the crossing and then put the ranges the wrong way round. The fix is the same as for the trucks: test one value. At five windows the first company charges 25 dollars and the second charges 55, so the first wins on small jobs.

61. Putting it together

Section

Part 7

62. Pattern: modelling a described situation

Pattern

Figure (svg): A flow diagram of five boxes: name the quantities and units, then find the rate and the initial value, then write the rule; below, answer by evaluating or solving, then state the domain and check the answer against it.

Four of the five boxes are algebra you already have. The last one is what turns algebra into a model.
  1. Name the quantities. Which is the input, which is the output, and in what units? Decide where input zero sits, and choose it so a given value lands there.
  2. Find the rate and the initial value. A rate is a change over a change, with its sign taken from the verb. Two data points give the rate; one point plus the rate gives the intercept.
  3. Write the rule. Slope-intercept when the value at zero is known, point-slope when it is not.
  4. Evaluate or solve. A given input means substitute; a given output means solve. Two competing plans mean set the models equal.
  5. State the domain, then check the answer against it. An answer outside the domain is not an answer, and an answer far outside the data is extrapolation, not prediction.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, pp. 232-233 — the eight-step strategy, condensed

This is the order to work in for any modelling problem in this chapter, and the diagram is worth a photograph. The first box is the one people skip and the last box is the one people forget.

Notice that the rate and the intercept are found in the same box, because they come from the same reading of the sentence, and that evaluating and solving sit in one box too, because the only difference between them is which of the two quantities the problem handed you.

The red arrow matters. When a final answer is unreasonable, the mistake is very rarely in the algebra: it is usually a rate with the wrong sign, or an input measured from the wrong starting point, or a quantity chosen as the output when it should have been the input. So the arrow goes all the way back to the first box, not to the one before.

63. Check: reading the rate from a sentence

Check

Check your understanding

A tank holds 200 litres and drains at 3 litres per minute. Which model and slope interpretation are both correct?

  • A. V(t) = 200 − 3t, slope −3 litres per minute (correct)
  • B. V(t) = 3t + 200, slope 3 litres per minute
  • C. V(t) = 200 − 3t, slope −3 minutes per litre
  • D. V(t) = 3t − 200, slope 3 litres per minute

Answer: A

Why: Draining lowers the volume, so the rate is negative, and the initial value is the 200 litres present at time zero. The units of the slope are output units over input units, which is litres per minute.

Why B tempts people
The sign is wrong. This model fills the tank, so the volume would pass 300 litres after a hundred seconds of draining.
Why C tempts people
The model is right but the units are upside down. Minutes per litre would be the rate of the inverse model, where volume is the input.
Why D tempts people
This starts the tank at minus 200 litres and fills it, which matches nothing in the sentence.

Two things are being tested at once here, the sign of the rate and the units of the slope, and three of the four options get exactly one of them right.

Draining lowers the volume, so the rate is negative, and the 200 litres present at the start is the value at time zero. The units are litres per minute because a slope is always output units divided by input units, and the output here is a volume.

The option with the correct model and reversed units is the one worth studying. Minutes per litre would be the slope of the inverse model, the one that takes a volume and returns how long it took to reach it. It is a perfectly good rate for a different question, which is why it reads plausibly.

64. Check: a reasonable domain

Check

Check your understanding

A newborn weighs 7.5 pounds and gains half a pound a month for its first year. What is the reasonable domain of the weight model?

  • A. all real numbers
  • B. from 0 to 12 months (correct)
  • C. from −15 to 12 months
  • D. from 7.5 to 13.5

Answer: B

Why: The domain is a set of inputs, and the description covers the first year only, so the inputs run from birth at zero months to twelve months.

Why A tempts people
That is the domain of the formula, not of the model. Negative months and ten-year-olds are not described by this sentence.
Why C tempts people
Minus 15 months is where the formula's horizontal intercept sits, which is exactly the input the reasonable domain is meant to exclude.
Why D tempts people
Those are weights, so that is the range. The domain is measured in months.

Before choosing, say what kind of object a domain is. It is a set of inputs, and here the input is measured in months, which rules out one of the four options immediately.

The description covers the baby's first year, so the reasonable domain runs from zero to twelve months. All real numbers is the formula's domain, not the model's, and the option that starts at minus fifteen months has taken the formula's horizontal intercept and treated it as a boundary of the situation, which is precisely backwards.

If you picked the option running from 7.5 to 13.5, you found the range instead. That is a good sign in one way, since those are the right two numbers, but keep the two words apart: domain is inputs, range is outputs.

65. Check: comparing two models

Check

Check your understanding

Two rental costs are K(d) = 0.59d + 20 and M(d) = 0.63d + 16, and they are equal at 100 miles. For which trips is Keep on Trucking cheaper?

  • A. fewer than 100 miles
  • B. more than 100 miles (correct)
  • C. exactly 100 miles
  • D. every trip, since 0.59 is less than 0.63

Answer: B

Why: Keep starts 4 dollars higher but gains 4 cents a mile on the other plan, so it needs 100 miles to make up the difference and is cheaper on anything longer.

Why A tempts people
Below 100 miles the extra 4 dollars up front has not been recovered yet, so Move It Your Way is still the cheaper one.
Why C tempts people
At exactly 100 miles both cost 79 dollars, so neither is cheaper.
Why D tempts people
The lower rate only wins once enough miles have been driven for it to overcome the higher up-front fee.

The crossing is given to you, so the only question is which side belongs to which company, and that is exactly the part that gets reversed under pressure.

Reason from the two numbers. Keep on Trucking starts 4 dollars more expensive and gains 4 cents a mile on the other plan. Gaining 4 cents a mile, it takes 100 miles to recover 4 dollars, and after that it is ahead. So Keep is cheaper on the longer trips.

The last option is the standard trap in every comparison problem: choosing the plan with the lower rate and ignoring the fee entirely. A lower rate only wins once enough input has accumulated to pay off the higher fee, and computing when that happens is the whole purpose of the exercise.

66. Explain the choice of input zero

Explain it to yourself

Discussion prompt

Two students model the same town. One writes the population as a function of the calendar year; the other writes it as a function of years since 2004. Both are correct. In two or three sentences, explain what changes between the two models and what does not, and why the second is usually preferred.

Write your two or three sentences before revealing. Being able to explain why a choice is free and still worth making carefully is a better test of understanding than any calculation on this slide.

The things that do not change are the rate and every prediction the model makes about a real date. Both students agree that the town grows by 380 people a year and that the 2013 population is 9,620. What changes is the intercept and what the input means, so the two models are not interchangeable: putting 2013 into the model built on years since 2004 gives an answer in the hundreds of thousands.

That is why the shift is worth making and why it must be written down. Say in words what your input measures, every time, and the reader knows which of the two models you are handing them.

67. Where the doughnut model must break

Counterexample

\[ C(x) = 0.25x + 25000 \]

Discussion prompt

This model says every doughnut costs exactly 25 cents to produce, for ever. Describe two different inputs at which it must stop describing the company, and say which feature of a linear function is responsible in each case.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 235 — Try It 2.14

Answer both halves before revealing, and try to name a feature of linear functions in each case rather than just telling a story about doughnuts.

At the high end the culprit is the constant rate of change. A linear model says the millionth doughnut costs exactly what the first one did, and no real business works that way: at some point you need another oven, another shift, another building, and the cost jumps. Real cost curves bend, and the linear model is only the straight piece near the middle.

At the low end the culprit is the unrestricted domain. The formula happily accepts minus a thousand doughnuts and returns a smaller cost. Nothing in the algebra forbids it; only the situation does. Both failures are ordinary and neither makes the model useless, so long as you state the stretch of production over which you are claiming it holds.

68. Exit ticket: the moose population

Exit ticket

In 1991 the moose population in a park was measured at 4,360. By 1999 it was 5,880. Assume the population changes linearly, and let the input be years since 1990.

Discussion prompt

Find the model, predict the population in 2003, and say whether that prediction is interpolation or extrapolation.

OpenStax, Precalculus, §2.3 Modeling with Linear Functions §2.3, p. 244 — Exercise 202

Work the whole thing before revealing, and pay attention to the input definition, because this problem is set up to punish carelessness with it: the data are from 1991 and 1999, but the input counts years since 1990.

That means the data points are at inputs 1 and 9, not 0 and 8, so neither of them is the intercept and you have to compute it. The rate is 1,520 moose over 8 years, which is 190 a year, and running back one year from 4,360 gives an intercept of 4,170. Then 2003 is input 13, giving 6,640 moose.

The last part is the one that shows whether the second half of this lesson landed. The data end in 1999 and the prediction is for 2003, so it is extrapolation, four years beyond the last measurement. That is a mild extrapolation and probably reasonable, but it is still a different kind of claim from interpolating between the two measurements, and saying so is part of the answer.

69. Recap

Recap

the questionwhat to do about it
which quantity is the input?the one you count off or choose; the other one depends on it
what is the rate?change in output over change in input, with units, and the sign taken from the verb
what is the initial value?the output when the input is zero, which may not be one of the given points
where does the input start?wherever you like; choose it so a given value becomes the intercept
what is the domain?the inputs the situation allows, not the ones the formula accepts
can I trust this answer?inside the data it is interpolation; far outside it, the model has broken down
two plans to compare?set the models equal; the crossing input splits the answer into two ranges

\[ f(x) = mx + b, \qquad x \in \text{[what the situation allows]} \]

Section 2.4 takes the next step: real data never sit exactly on a line, so it asks how to fit the best line through a scatter of points, and how to judge how good the fit is.

OpenStax Algebra and Trigonometry 2e, §4.2 Modeling with Linear Functions §4.2 — the same section in Algebra and Trigonometry

Read the table as a list of the questions a modelling problem can ask, because between them they cover nearly every exercise in the section. The algebra behind them is all from Sections 2.1 and 2.2; what this lesson added was the reading at the front and the judgement at the back.

Two habits are worth carrying out of here. Write the units on every rate, which catches upside-down slopes and misidentified inputs before they reach an answer. And write the reasonable domain on the same line as the model, which catches answers the situation cannot support.

Section 2.4 removes the assumption that has been quietly holding all of this together. Real measurements do not sit exactly on a line, so instead of a model through two chosen points you will fit a line to a whole scatter of data, and you will need a way to say how good the fit is. Interpolation, extrapolation and model breakdown all return there, with more to say.

Sources

  1. OpenStax, Precalculus, §2.3 Modeling with Linear Functions
  2. OpenStax Algebra and Trigonometry 2e, §4.2 Modeling with Linear Functions

Want this taught 1-on-1? Alexander tutors Precalculus — $55/session, free consultation.

Book on Wyzant · Text (657) 465-8108