5.2 The Uniform Distribution

Section 5.1's flat density, named and given parameters. The uniform distribution is a continuous distribution concerned with events that are equally likely to occur across an interval, written X follows U of a and b, where a is the lowest value the variable can take and b the highest. Its density is one over b minus a, a height forced by the requirement that a rectangle of that width enclose an area of one. Because the region is always a rectangle, every question reduces to arithmetic on a base and a height: a probability is base times height, the mean is the midpoint of a and b, the standard deviation is the width divided by the square root of twelve, a percentile is found by setting the area equal to the given proportion and solving for the boundary, and a conditional probability is found by redrawing the density on the reduced interval.

Subject: Statistics · 68 slides · symbolic lesson

Open the interactive version of this deck

What this lesson covers

The lesson, slide by slide

1. The Uniform Distribution

Title

Statistics · Chapter 5 — Continuous Random Variables

Equally likely across an interval: every question is a rectangle

2. What this lesson gives you

Objectives

Section 5.1 showed you that for a continuous random variable, probability is area under a density curve. This section takes the simplest density there is, a flat one, and gives it a name and two parameters. Because the region is always a rectangle, every question turns into arithmetic on a base and a height, and you can draw the answer before you compute it.

The arithmetic in this section is easy. What costs marks is the setup: the wrong height, a base that starts at zero when the interval does not, a percentile read from the wrong side, a condition ignored. Every one of those is visible in a sketch, so draw the rectangle every time.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, pp. 295-302 — the section this deck covers

3. Before anything new: area under a flat line

Warm-up

Section 5.1 ended with a flat density. Recall how you found a probability from it, on the book's Try It 5.1.

Discussion prompt

A density is f of x equal to one eighth for x between 0 and 8. Sketch it, check that it encloses an area of one, and find the probability that x is between 2.5 and 7.5.

Hold on to two facts from that answer. The total area had to be exactly one, and it was, because eight times one eighth is one. And the probability was just the width of the strip times the height, with no curve to fight. Both facts are the whole of this section.

Ask yourself one more question: where did the height one eighth come from? Section 5.1 handed it to you. Here you will see that you never need to be handed it, because the width of the interval forces it.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.1, p. 295 — Try It 5.1

4. Naming the distribution

Section

Part 1

5. The data behind the assumption

Picture it

Example 5.2 starts with real data: 55 smiling times, in seconds, of an eight-week-old baby. The shortest is 0.7 seconds and the longest 22.8. Before any model, look at how the times spread across that range.

Figure (svg): A histogram of the 55 smiling times in five bins 4.6 seconds wide from 0 to 23, with counts 8, 12, 14, 11, 10. A dashed yellow line at 11 marks what each bin would hold if the times were spread perfectly evenly. Orange ticks under the axis mark each of the 55 values.

Five equal bins, and each holds between 8 and 14 of the 55 times. A perfectly flat spread would put 11 in each. The orange rug shows the times scattered across the whole interval with no pile-up anywhere.

Cut the range from 0 to 23 into five equal bins. The counts are 8, 12, 14, 11, 10. If every smiling time from 0 to 23 seconds were equally likely, each bin would expect one fifth of the 55 values, which is 11. The real counts wander around 11 without piling up anywhere.

\[ \text{expected per bin} = \frac{55}{5} = 11 \]

That is the evidence for the book's assumption that the smiling times follow a uniform distribution between 0 and 23 seconds: any smiling time in that range is as likely as any other. The histogram is the empirical distribution; the flat line at 11 is what the theoretical one predicts.

Notice what the model does not claim. It does not say every bin will hold exactly 11; a sample of 55 is always lumpy. It says there is no region of the interval that smiles prefer. The rug of orange ticks under the axis shows exactly that: values scattered from one end to the other.

The book reports a sample mean of 11.65 seconds and a sample standard deviation of 6.08. Recomputed from Table 5.1, they are 11.653 and 6.081, so the printed values check. You will compare them with the model's in Part 3.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 295 — Example 5.2 and Table 5.1

6. Two numbers name the distribution

Concept

A uniform distribution is concerned with events that are equally likely to occur across an interval. Once you know where the interval starts and where it stops, you know everything about it.

uniform distribution — X ~ U(a, b), read X is uniformly distributed between a and b, where a is the lowest value X can take and b the highest.

\[ X \sim U(a, b), \qquad a = \text{lowest value of } x, \quad b = \text{highest value of } x \]

For the smiling times, the random variable is the length of one baby's smile, and the interval runs from 0 to 23 seconds.

\[ X = \text{length, in seconds, of an eight-week-old baby's smile} \]

\[ X \sim U(0, 23) \]

Always define X in words first, with its units. The letters a and b are values of X, so they carry the same units: 0 seconds and 23 seconds. Writing U of 0 and 23 without saying seconds of what leaves the reader to guess what the numbers measure.

The order matters too. The smaller number goes first: U of 23 and 0 is not a different distribution, it is a mistake, and it will give you a negative width later. The book also warns you to notice whether the endpoints are included; a later slide shows why, for a continuous variable, that turns out not to change any probability.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 295 — the notation X ~ U(a, b)

7. The height is forced by an area of one

Concept

A density must enclose a total area of exactly one, because the probability that X takes some value in its interval is one. For a uniform distribution the region is a rectangle, and that single fact fixes its height.

\[ \text{area} = \text{base} \times \text{height} = 1 \]

\[ (b - a) \times \text{height} = 1 \]

\[ \text{height} = \frac{1}{b - a} \]

Figure (svg): Three small graphs on identical axes. The first is a square from 0 to 1 of height 1. The second is a rectangle from 0 to 2 of height one half. The third is a long low rectangle from 0 to 4 of height one quarter. Each is labelled area equals 1.

Stretch the interval and the rectangle must flatten: the total area is locked at one, so the height is one over the width.

The three rectangles have widths 1, 2 and 4. Each has area one, so the heights are 1, one half and one quarter. A wide interval means a low rectangle; a narrow one means a tall rectangle. You never have to memorise the height, only remember that the area is one and divide.

One consequence surprises people. The height is not a probability, and it can be bigger than one: a uniform distribution on an interval half a unit wide has height 2. What cannot exceed one is an area. That is why the next slide calls the height a density rather than a probability.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 295 — the probability density function

8. Reading the density piece by piece

Notation

Here is the density exactly as the book writes it. Every part of it has a job, and the part on the right is as important as the fraction.

Annotate

On: \( f(x) = \frac{1}{b - a} \quad \text{for } a \le x \le b \)

  • The density: the height of the graph above the value x. Areas under it, not heights, are probabilities.
  • One divided by the width of the interval, the only height that makes the rectangle's area one. It does not depend on x, which is why the graph is flat.
  • The width of the interval, in the units of X. It is the lowest value subtracted from the highest, never just b.
  • Where the formula holds. Outside the interval the density is zero: no smile is shorter than 0 seconds or longer than 23 in this model.

For the smiling times the width is 23 minus 0, so the height is one twenty-third, about 0.0435 per second. The words per second are a reminder that a density is probability per unit of x; multiply by a width in seconds and you get a probability with no units.

\[ f(x) = \frac{1}{23 - 0} = \frac{1}{23} \quad \text{for } 0 \le x \le 23 \]

The domain condition is where half of the section's mistakes begin. A base that starts at zero when the interval starts at 1.5, or a base that runs past b, both come from forgetting that the formula only holds between a and b.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 295 — f(x) = 1/(b − a) for a ≤ x ≤ b

9. Example 5.2: writing the smiling-time distribution

Worked example

The smiling times of an eight-week-old baby are assumed to be uniform between 0 and 23 seconds, inclusive. Name the random variable, write the distribution, and write its density.

Define X in words, with units

Why: the quantity that varies from smile to smile.

\[ X = \text{length, in seconds, of an eight-week-old baby's smile} \]

Read off a and b

Why: the shortest and the longest possible smile.

\[ a = 0, \quad b = 23 \]

Write the distribution

Why: lowest value first.

\[ X \sim U(0, 23) \]

Find the width

Why: b minus a.

\[ b - a = 23 - 0 = 23 \]

Write the density

Why: one over the width, on the interval.

\[ f(x) = \frac{1}{23} \quad \text{for } 0 \le x \le 23 \]

Figure (svg): The graph of f of x equals 1 over 23 from 0 to 23 seconds: a flat rectangle, shaded, with the left end labelled a equals 0, the right end labelled b equals 23, and the height labelled 1 over 23. Inside, the label reads area equals 23 times 1 over 23 equals 1.

The whole story in one picture: a flat top over the possible smiling times, as high as it needs to be to enclose an area of exactly one.

The sketch is part of the answer. Label the horizontal axis with the variable and its units, mark a and b, and mark the height. Every later question about smiles will shade some part of this rectangle.

Why 23 and not 22.8, the longest smile in the sample? The model describes every smile an eight-week-old might produce, not just these 55, and the book chooses a round upper limit just beyond the data. The choice of a and b is a modelling decision, and it should be stated, as the book does, as an assumption.

Check that the area is one

Why: base times height over the whole interval must give a total probability of one.

\[ (23 - 0)\left(\frac{1}{23}\right) = 1 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 295 — Example 5.2; the book writes the cases of X and x the other way round in the domain, and the deck writes 0 ≤ x ≤ 23

10. Try It 5.6(a)-(c): the oil change, an interval not starting at zero

Worked example

The time a service technician needs to change the oil in a car is uniformly distributed between 11 and 21 minutes. Write X in words, write the distribution, and graph it.

Write X in words

Why: the book's part (a) asks for exactly this sentence.

\[ X = \text{the time, in minutes, needed to change the oil on a car} \]

Read off a and b

Why: no oil change is quicker than 11 minutes or slower than 21.

\[ a = 11, \quad b = 21 \]

Write the distribution

Why: part (b).

\[ X \sim U(11, 21) \]

Find the width

Why: the highest value minus the lowest.

\[ b - a = 21 - 11 = 10 \]

Write the height

Why: one over the width.

\[ f(x) = \frac{1}{10} = 0.1 \quad \text{for } 11 \le x \le 21 \]

Figure (svg): A graph with the x-axis from 0 to 24 minutes. Nothing is drawn from 0 to 11. A rectangle of height 1 over 10 runs from 11 to 21 minutes. The gap from 0 to 11 is labelled f of x equals 0 here.

The rectangle starts at 11, not at 0. Its width is 10 minutes, so its height is one tenth, and below 11 the density is zero: no oil change is that fast.

Part (c) asks for the graph. Drawing the axis from zero makes a point you will need later: the rectangle begins at 11, and between 0 and 11 the density is zero. A base measured from 0 would count ten minutes in which no oil change can happen.

Notice also that the height has nothing to do with 11 or 21 on their own. Any ten-minute interval, 0 to 10 or 100 to 110, gives the same height of one tenth. The position of the rectangle is set by a and b; its shape is set only by the distance between them.

Check the area and the height

Why: ten minutes wide times one tenth tall is one; a height of one over 21 would give 10 over 21, less than one.

\[ (21 - 11)(0.1) = 1 \quad \text{but} \quad (10)\left(\tfrac{1}{21}\right) \approx 0.476 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 302 — Try It 5.6, parts (a) to (c)

11. Uniform, or not?

Sorting

A uniform model claims that no part of the interval is more likely than any other. For each quantity, ask whether that is believable: is there a range of values it favours?

Sort into buckets

Sort each quantity by whether a uniform model is reasonable.

uniform is reasonable
your wait for a train that leaves every 10 minutes, when you arrive at a random moment; the seconds shown on a clock's second hand when you glance at it at a random moment; the point along a thin rod where it snaps, if every point is equally weak
not uniform: some values are favoured
the heights of adult women; the time until the next earthquake; the amount a customer spends on one trip to the supermarket
u
Each has a clear lowest and highest value, and nothing makes one part of the range more likely than another: a random moment is as likely to fall early in the ten minutes as late.
n
Each piles up somewhere. Heights cluster near the middle; short waits for an earthquake and small shopping bills are more common than long waits and large bills.

The three uniform cases share a structure: a random moment or a random point falls into a fixed range with no preference. The train is the standard example. If you do not know the timetable, you are as likely to arrive one minute after a departure as nine minutes after.

The earthquake wait and the supermarket bill are the examples the book uses at the start of Section 5.3 for the exponential distribution, where small values are more common than large ones. Heights belong to the normal distribution of Chapter 6. A model has to match the shape, not just the range.

If you sorted the heights as uniform because heights do lie between some lowest and highest value, notice that having a range is not enough. Every variable has a range. Uniform means flat across it.

12. One value has probability zero, so endpoints do not matter

Concept

The book warns you to note whether the data are inclusive or exclusive of the endpoints. For a continuous variable, including or excluding a single value changes nothing, and the picture shows why.

Figure (svg): The uniform density of height 0.05 from 0 to 20. A narrow shaded strip from 9.5 to 10.5 is labelled width 1, area 0.05. A single bright vertical line at x equals 15 is labelled width 0, area 0.

Probability is area. The strip has some width and so some area; the single value 15 is a line with no width, so its probability is zero.

Take a uniform variable on 0 to 20, with height 0.05. The probability of a narrow strip, one unit wide, is one times 0.05. Shrink the strip to the single value 15 and its width is zero.

\[ P(x = 15) = (\text{base})(\text{height}) = (0)(0.05) = 0 \]

So adding or removing one point adds or removes zero area. Strictly less than and less than or equal to give the same answer, and so do strictly between and between inclusive.

\[ P(x < 12.5) = P(x \le 12.5), \qquad P(2 < x < 18) = P(2 \le x \le 18) \]

This is a real difference from Chapter 4. For a count, fewer than five and at most five differ by a whole bar. For a continuous measurement they are identical. What does still matter is which side of the boundary the question means: more than 12 and less than 12 are different areas, however you treat the point 12 itself.

That is also why the book can call P of X less than x the cumulative distribution function, the CDF, and write it with either symbol. Part 2 draws it.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.1, p. 294 — P(x = 15) = 0 and the CDF

13. How tall is the furnace density?

Estimation

Example 5.6 says the time a repair technician needs to fix a furnace is uniformly distributed between 1.5 and 4 hours. Before writing any formula, estimate the height of its density from the width alone.

Predict first

What is the height of the density for repair times uniform between 1.5 and 4 hours?

  • 0.4
  • 0.25
  • 2.5
  • 1

Correct: 0.4

Why: The width is 4 minus 1.5, which is 2.5 hours, and the height is one over the width, so one over 2.5, which is 0.4. Using only b gives one quarter; the width itself is 2.5; and a height of 1 would make the area 2.5 instead of 1.

\[ \text{height} = \frac{1}{4 - 1.5} = \frac{1}{2.5} = 0.4 \]

A quick estimate is enough to rule out most of the options. The interval is a bit wider than 2 hours, so the height must be a bit less than one half. Only 0.4 and 0.25 survive that thought, and 0.25 comes from forgetting that the interval starts at 1.5.

Check any height by multiplying back: 2.5 times 0.4 is 1. The other candidates fail that check at once: 2.5 times 0.25 is 0.625, too little area, and 2.5 times 1 is 2.5, too much. This height, 0.4, is the number you will use in every furnace question in the deck.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 300 — Example 5.6(a)

14. Probabilities as rectangles

Section

Part 2

15. A probability is the interval's share

Concept

To find the probability that X lands between two values c and d, shade that part of the rectangle and find its area. The height is the same everywhere, so only the base changes.

\[ P(c < x < d) = (\text{base})(\text{height}) = (d - c)\left(\frac{1}{b - a}\right) \]

\[ P(c < x < d) = \frac{d - c}{b - a} \]

Read the second line as a sentence: the probability is the width you want as a share of the whole width. For the smiling times, any stretch of 2.3 seconds has probability one tenth, wherever it sits between 0 and 23. That is what equally likely across an interval means.

The formula holds only while c and d lie inside the interval. If a question reaches outside it, first cut the base back to the part inside, because the density is zero beyond a and b. For the smiling times, the probability of a smile shorter than 30 seconds is one, not 30 over 23.

\[ P(x < 30) = P(x < 23) = 1 \quad \text{not} \quad \frac{30}{23} \]

An answer bigger than one is always a sign that the base ran outside the rectangle. An answer that seems too small often means the base was measured from 0 when the interval starts somewhere else. Both are caught by a sketch.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 296 — P(2 < x < 18) = (base)(height)

16. Example 5.3(a): a smile between 2 and 18 seconds

Worked example

Refer to the smiling times, X ~ U(0, 23). What is the probability that a randomly chosen eight-week-old baby smiles between 2 and 18 seconds?

Write the event

Why: strictly between 2 and 18; the endpoints would add nothing.

\[ P(2 < x < 18) \]

Find the base

Why: the width of the interval asked about.

\[ 18 - 2 = 16 \]

Recall the height

Why: one over the width of the whole interval.

\[ \text{height} = \frac{1}{23} \]

Multiply base by height

Why: the area of the shaded rectangle.

\[ P(2 < x < 18) = (16)\left(\frac{1}{23}\right) \]

Simplify

Why: sixteen twenty-thirds.

\[ = \frac{16}{23} \approx 0.6957 \]

Figure (svg): The uniform density of height 1 over 23 from 0 to 23. The part from 2 to 18 is shaded, labelled base 16 and area 16 over 23, about 0.6957. The unshaded slivers at each end are narrow.

The shaded rectangle is the probability. Its base is the width of the interval asked about, 16 seconds; its height is the same one over 23 as everywhere else.

In context: about 70 percent of smiles of eight-week-old babies last between 2 and 18 seconds. The shaded base covers 16 of the 23 seconds, a bit over two thirds of the width, which matches.

How does the model compare with the data? In Table 5.1, 43 of the 55 smiles, about 78 percent, lie between 2 and 18 seconds, a little more than the model's 70 percent. The real smiles cluster slightly toward the middle, the same pattern the standard deviations will show in Part 3.

Check with the complement

Why: the two unshaded slivers, 0 to 2 and 18 to 23, must make up the rest of the area.

\[ \frac{2}{23} + \frac{5}{23} = \frac{7}{23}, \qquad \frac{16}{23} + \frac{7}{23} = 1 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 296 — Example 5.3(a)

17. Example 5.4(a): waiting fewer than 12.5 minutes for a bus

Worked example

The time a person must wait for a bus is uniformly distributed between 0 and 15 minutes, inclusive. What is the probability that a person waits fewer than 12.5 minutes?

Define X and write the distribution

Why: the wait, from 0 to 15 minutes.

\[ X = \text{minutes a person waits}, \quad X \sim U(0, 15) \]

Write the density

Why: one over the width of 15 minutes.

\[ f(x) = \frac{1}{15} \quad \text{for } 0 \le x \le 15 \]

Write the event

Why: fewer than 12.5 minutes; the rectangle starts at 0, so the base does too.

\[ P(x < 12.5) \]

Base times height

Why: the base runs from 0 to 12.5.

\[ P(x < 12.5) = (12.5 - 0)\left(\frac{1}{15}\right) \]

Simplify

Why: twelve and a half fifteenths.

\[ = \frac{12.5}{15} \approx 0.8333 \]

Figure (svg): The uniform density of height 1 over 15 from 0 to 15 minutes. The part from 0 to 12.5 is shaded and labelled area 12.5 over 15, about 0.8333. The unshaded piece from 12.5 to 15 is labelled 0.1667.

Most of the rectangle is shaded: fewer than 12.5 minutes covers five sixths of the possible waits. The leftover sliver is the chance of a longer wait.

In context: about 83 percent of the time, a person waits fewer than 12.5 minutes for the bus. Here a is 0, so the base happens to start at 0. That will not be true for the furnace, and the book makes a point of it there.

Fewer than and at most give the same answer here, as the endpoint slide showed. So the book's words, fewer than 12.5 minutes, and a rider's words, no more than 12.5 minutes, describe the same shaded region and the same 0.8333.

Check with the unshaded sliver

Why: the chance of waiting at least 12.5 minutes is the leftover 2.5 minutes of width, and the two must add to one.

\[ (15 - 12.5)\left(\tfrac{1}{15}\right) \approx 0.1667, \quad 0.8333 + 0.1667 = 1 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 298 — Example 5.4(a)

18. Area to the left, and area to the right

Concept

The book names one probability in particular: the area to the left of a value x, which it calls the cumulative distribution function, or CDF. For a uniform distribution you can write it straight from the rectangle.

\[ P(X < x) = (x - a)\left(\frac{1}{b - a}\right) = \frac{x - a}{b - a} \quad \text{for } a \le x \le b \]

Figure (svg): A graph of the probability that X is less than x, for the smiling times. It is 0 up to x equals 0, rises in a straight line to 1 at x equals 23, and stays at 1 after. A dot at x equals 18 sits at height 0.7826, with dashed guides to both axes.

The area to the left grows at a steady rate, one twenty-third per second, because the density is flat. So the cumulative graph is a straight ramp from 0 at a to 1 at b.

The graph of the CDF is a straight ramp. Each extra second of x adds one twenty-third of area, the density's height, so the ramp rises at that constant rate from 0 at a to 1 at b. At 18 seconds it has reached 18 over 23, about 0.7826.

The area to the right is whatever is left of the total of one. That is the book's rule for a greater-than question on a continuous distribution.

\[ P(X > x) = 1 - P(X < x) \]

For a uniform distribution you can also find a right-hand area directly, as the base from x up to b times the height. Both routes must agree, which makes the complement a free check on every answer. The ramp will earn its place in Part 4, where you read it backwards.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.1, p. 294 — the CDF and P(X > x) = 1 − P(X < x)

19. Example 5.6(a),(b): furnace repairs longer than 2 and shorter than 3 hours

Worked example

The time to fix a furnace is uniformly distributed between 1.5 and 4 hours. Find the probability that a randomly selected repair takes more than two hours, and the probability that it takes less than three.

Write the distribution and the height

Why: the width is 2.5 hours.

\[ X \sim U(1.5, 4), \quad f(x) = \frac{1}{4 - 1.5} = 0.4 \]

Part (a): the base for more than 2

Why: from 2 up to the top of the interval, 4.

\[ P(x > 2) = (4 - 2)(0.4) \]

Multiply

Why: two hours of base.

\[ = 0.8 \]

Part (b): the base for less than 3

Why: from the bottom of the interval, 1.5, up to 3; not from 0.

\[ P(x < 3) = (3 - 1.5)(0.4) \]

Multiply

Why: an hour and a half of base.

\[ = 0.6 \]

Figure (svg): Two graphs of the uniform density of height 0.4 from 1.5 to 4 hours, drawn with the x-axis starting at 0. On the left, the region from 2 to 4 is shaded, area 0.8. On the right, the region from 1.5 to 3 is shaded, area 0.6.

Both shaded bases stop at the ends of the rectangle. On the right the base starts at 1.5, not 0, because no repair takes less than an hour and a half.

The book stresses the second base: the shaded area starts at 1.5 rather than at 0, because X cannot be less than 1.5. In context: 80 percent of repairs take more than two hours, and 60 percent take less than three.

The two answers add to 1.4, which is not a mistake: the events overlap. Repairs between two and three hours belong to both, and that overlap is (3 minus 2) times 0.4, which is 0.4. Two probabilities about one variable need not add to one unless their events are complements.

Check part (a) with the complement

Why: less than 2 is the base from 1.5 to 2, and it must make up the rest.

\[ (2 - 1.5)(0.4) = 0.2, \quad 0.2 + 0.8 = 1 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, pp. 300-301 — Example 5.6(a) and (b), Figures 5.18 and 5.19

20. Four attempts at P(x < 3) for the furnace

Error analysis

Four students computed the probability that a furnace repair, uniform between 1.5 and 4 hours, takes less than three hours. One is right. Find the fault in each of the others.

Annotate

On: \( (3)(0.4) = 1.2 \quad (3)\left(\tfrac{1}{4}\right) = 0.75 \quad (3 - 1.5)(0.4) = 0.6 \)

  • The base was measured from 0, counting an hour and a half in which no repair can finish. The answer is bigger than one, which no probability can be: that alone flags it.
  • The height is one over b instead of one over the width, and the base starts at 0. The two slips are the same mistake, forgetting a, made twice.
  • Correct. The base runs from where the rectangle starts, 1.5, to 3; the height is one over 2.5.
  • Another student wrote (3 − 1.5)(1/4) = 0.375, right base but height one over b. The shaded area would then be less than it should be, and the whole rectangle would have area 0.625.

Every wrong line forgets that the interval starts at 1.5. When a is zero, as for the smiles and the bus, the base from 0 and the base from a are the same, and the slip is invisible. The furnace is where it shows.

Two quick checks catch all three. First, a probability above one is impossible. Second, the whole rectangle must have area one: with height one quarter over a width of 2.5 it has area 0.625, so that height cannot be right.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 301 — Example 5.6(b): the shaded area starts at x = 1.5 rather than at x = 0

21. Try It 5.4(a)-(d): a baseball season between 447 and 521 hours

Worked example

The total duration of baseball games in a major league season is uniformly distributed between 447 and 521 hours, inclusive. Find a and b and say what they represent, write the distribution, and find the probability that a team's season lasts between 480 and 500 hours.

Name a and b

Why: the shortest and longest possible season total, in hours.

\[ a = 447, \quad b = 521 \]

Write the distribution

Why: X is the total hours of games for a team in one season.

\[ X \sim U(447, 521) \]

Find the width and height

Why: 74 hours wide.

\[ b - a = 74, \quad f(x) = \frac{1}{74} \]

Write the event and its base

Why: 480 to 500 lies inside the interval.

\[ P(480 < x < 500) = (500 - 480)\left(\frac{1}{74}\right) \]

Simplify

Why: twenty seventy-fourths.

\[ = \frac{20}{74} \approx 0.2703 \]

Figure (svg): The uniform density of height 1 over 74 from 447 to 521 hours, on an x-axis from 440 to 530. The region from 480 to 500 is shaded and labelled 20 over 74, about 0.2703.

The axis does not start at zero, and it does not need to. Only the width of the interval matters: 74 hours in all, of which the shaded 20 is a little over a quarter.

In context: about 27 percent of teams' seasons last between 480 and 500 hours. The large numbers change nothing: only differences between them enter the calculation. The book's parts (c) and (e), the mean and the 65th percentile, return later in the deck.

Check the share

Why: 20 hours out of a 74-hour range should be a little over a quarter of the area.

\[ \frac{20}{74} \approx 0.27 \quad \text{and} \quad \frac{1}{4} = 0.25 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 299 — Try It 5.4, parts (a), (b) and (d)

22. Which smiling interval is more likely?

Prediction

For the smiling times, uniform on 0 to 23 seconds, compare two intervals before you compute anything: a smile between 2 and 5 seconds, and a smile between 18 and 21 seconds.

Predict first

Which is more likely for a smile uniform between 0 and 23 seconds?

  • between 2 and 5 seconds
  • between 18 and 21 seconds
  • they are equally likely
  • it depends on the sample

Correct: they are equally likely

Why: Both intervals are 3 seconds wide and lie inside the rectangle, so both have base 3 and height one over 23. Their probabilities are both three twenty-thirds, about 0.13. Under a uniform model only the width of an interval matters, not where it sits.

\[ P(2 < x < 5) = P(18 < x < 21) = \frac{3}{23} \approx 0.1304 \]

If you felt that short smiles or long smiles should be more common, that feeling is exactly what the uniform model rules out. Real data might favour one end, and the histogram of Table 5.1 is lumpy, but the model has no preference.

This is the most useful single test of whether a uniform model fits. If you believe two equal-width intervals really do differ in probability, the model is wrong for your situation, however convenient its rectangles are.

23. Try It 5.3: a probability on U(0, 20)

Fill the middle

A distribution is given as X ~ U(0, 20). Fill in the base and the answer for the probability that x is between 2 and 18.

Fill in the blanks

P(2 < x < 18) = (16)\left(\frac0.8___\right) = ___

Why: The base is the width of the interval asked about, 18 minus 2, which is 16, and the height is one over 20. Sixteen twentieths is four fifths, 0.8.

Compare it with Example 5.3(a): the same interval, 2 to 18, on a slightly wider rectangle. There the answer was 16 over 23, about 0.70; here it is 16 over 20, 0.80. The same base covers a bigger share of a narrower interval.

If you wrote 18 in the first blank, you measured the base from 0 to 18, which answers the probability of less than 18. That is a different question, and its answer, 0.9, is also the clue to the next part of Try It 5.3, the 90th percentile, which Part 4 takes up.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 298 — Try It 5.3

24. The mean and standard deviation

Section

Part 3

25. The mean is the midpoint

Intuition

The mean of a distribution is its balance point: if you cut the density out of card, it is where the card would balance on a finger. A flat rectangle balances exactly in its middle.

Figure (svg): The bus-wait rectangle from 0 to 15 drawn as a flat slab resting on a triangular fulcrum at 7.5. Arrows show equal lengths of slab on each side, each labelled 7.5 minutes of area. Above the fulcrum: mu equals 0 plus 15 over 2 equals 7.5.

A flat slab balances at its centre. Every wait of 7.5 minus some amount is matched by an equally likely wait of 7.5 plus the same amount, so the average lands in the middle.

For the bus waits, uniform on 0 to 15 minutes, every wait some amount below 7.5 is matched by an equally likely wait the same amount above it: 7 with 8, 2 with 13, 0.5 with 14.5. The pairs average to 7.5, so the whole distribution does.

\[ \mu = \frac{a + b}{2} \]

\[ \mu = \frac{0 + 15}{2} = 7.5 \text{ minutes} \]

On average, a person waits 7.5 minutes for the bus. The formula is the ordinary average of the two endpoints, which is the midpoint of the interval. That is also why the mean of a uniform distribution is also its median: half the area lies on each side of the middle.

One caution for when the interval does not start at zero. The mean of the furnace times is 1.5 plus 4 over 2, which is 2.75 hours, not 4 over 2. Adding a before halving is what places the midpoint inside the interval.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, pp. 296, 298 — μ = (a + b)/2, and Example 5.4(b)

26. Where the twelve comes from

Concept

The book gives the standard deviation of a uniform distribution as the width divided by the square root of twelve. You can see where the twelve comes from without calculus, by treating the rectangle as many equally spaced, equally likely values.

Put n equally spaced values from 0 to 1 and give each probability one over n. With two values, 0 and 1, the variance is one quarter. With more values filling in the gaps, the variance settles down. Here are the values computed in node:

values from 0 to 1variancestandard deviation
20.25000.5000
30.16670.4082
50.12500.3536
110.10000.3162
1010.08500.2915
10010.08350.2890

As the values fill the whole interval, the variance approaches 0.0833, which is one twelfth, and the standard deviation approaches the square root of one twelfth, about 0.2887. That limit is the uniform distribution on an interval of width one.

\[ \sigma^{2} \to \frac{1}{12}, \qquad \sigma \to \frac{1}{\sqrt{12}} \approx 0.2887 \]

An interval of width b minus a is the same picture stretched. Stretching every value by a factor stretches every distance from the mean by that factor, so the standard deviation scales with the width.

\[ \sigma = \frac{b - a}{\sqrt{12}} = \sqrt{\frac{(b - a)^{2}}{12}} \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 296 — the formula for the theoretical standard deviation

27. The spread depends only on the width

Intuition

Look at what the standard deviation formula leaves out: a and b appear only as their difference. Slide the interval along the axis and the spread does not change; stretch it and the spread stretches in proportion.

Figure (svg): A straight line through the origin: standard deviation against the width of the interval, with slope 1 over root 12, about 0.2887. Dots on the line mark the furnace at width 2.5, the oil change at 10, the bus at 15 and the smiling times at 23.

Double the width and you double the spread. Where the interval sits on the axis plays no part: only b minus a appears.

Every uniform distribution in this deck sits on the one line, standard deviation equals width times 0.2887. The furnace, only 2.5 hours wide, has a standard deviation of about 0.72 hours; the smiling times, 23 seconds wide, about 6.64 seconds. The baseball season, 74 hours wide, is off the right of the graph at about 21.36 hours.

It helps to know what one standard deviation covers here. The band from one standard deviation below the mean to one above has width 2 over the square root of 12 times the width, so it holds a fixed share of the rectangle.

\[ \frac{2(b - a)/\sqrt{12}}{b - a} = \frac{2}{\sqrt{12}} \approx 0.577 \]

About 58 percent of a uniform distribution lies within one standard deviation of the mean, whatever the interval. That is less than the roughly 68 percent you will meet for the bell-shaped normal distribution, because a flat distribution has more of its area out near the ends.

28. Reading the standard deviation formula

Notation

The book writes the standard deviation under a square root. Read each piece, and notice which operation turns a variance back into a spread.

Annotate

On: \( \sigma = \sqrt{\frac{(b - a)^{2}}{12}} \)

  • The theoretical standard deviation: the typical distance of a value from the mean, in the units of X.
  • The width squared. Squaring changes the units to squared units, seconds squared for the smiles.
  • The constant for a flat shape: the variance of a uniform distribution of width one is one twelfth.
  • Undoes the squaring, so σ comes back in the units of X. Stop before it and you have the variance, not the standard deviation.

The quantity under the root is the variance. For the bus it is 15 squared over 12, which is 18.75 minutes squared; the square root brings it back to about 4.33 minutes. The two numbers are both correct answers, but to two different questions.

\[ \sigma^{2} = \frac{15^{2}}{12} = 18.75, \qquad \sigma = \sqrt{18.75} \approx 4.33 \]

Because the square root of a square is the original number, the form on the previous slides, width over root twelve, is the same thing written more simply. Use whichever you find easier to type into a calculator; both give identical values.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 296 — σ = √((b − a)²/12)

29. Example 5.2: the theoretical mean and standard deviation of the smiles

Worked example

For the smiling times, X ~ U(0, 23), compute the theoretical mean and standard deviation, and compare them with the sample mean and standard deviation of Table 5.1.

The mean is the midpoint

Why: average the endpoints.

\[ \mu = \frac{a + b}{2} = \frac{0 + 23}{2} \]

Simplify

Why: half of 23.

\[ \mu = 11.50 \text{ seconds} \]

Square the width and divide by 12

Why: the variance first.

\[ \sigma^{2} = \frac{(23 - 0)^{2}}{12} = \frac{529}{12} \approx 44.08 \]

Take the square root

Why: back to seconds.

\[ \sigma = \sqrt{44.08} \approx 6.64 \text{ seconds} \]

Figure (svg): The uniform density of height 1 over 23 from 0 to 23, lightly shaded. Two double-headed arrows are drawn inside it. The upper, yellow, runs from 4.86 to 18.14, labelled theory: 11.50 plus or minus 6.64. The lower, orange, runs from 5.57 to 17.73, labelled sample: 11.65 plus or minus 6.08.

The centres almost coincide. The sample's spread is a little narrower than the model's, because the real times thin out slightly near 0 and 23.

The sample of 55 smiles has mean 11.65 and standard deviation 6.08, recomputed from the data. The centres are within a fifth of a second of each other. The sample's spread is smaller than the model's: in the data, 37 of the 55 times, about 67 percent, lie within one theoretical standard deviation of 11.5, against the 58 percent the flat model predicts. The real smiles thin out a little near the ends.

That difference is modest, and the book's judgement is that the theoretical values are close to the sample values. The uniform model is an idealisation that describes these data well, not a claim that they are perfectly flat.

Check the standard deviation another way

Why: width over root twelve must give the same value.

\[ \frac{23}{\sqrt{12}} = \frac{23}{3.4641} \approx 6.64 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 296 — Example 5.2, theoretical mean and standard deviation

30. Try It 5.2: the fish caught on 35 charter boats

Worked example

The total weight of fish, to the nearest pound, caught on each of 35 charter boats on one summer day is modelled as uniform, with all values from 0 to 14 pounds equally likely. The sample mean is 7.9 and the sample standard deviation 4.33. State a and b, write the distribution, and find the theoretical mean and standard deviation.

Figure (svg): A dot plot of the 35 fish weights from 0 to 14 pounds, one dot per boat, stacked. The tallest stacks, five dots each, are at 4, 10 and 11 pounds; there are no dots at 8. A dashed yellow line at about 2.3 dots marks a perfectly even spread. Yellow and orange markers under the axis show the theoretical mean 7 and the sample mean 7.86.

Thirty-five values spread over fifteen possible weights, lumpy as small samples are, but with no region the boats avoid. The model's centre, 7, sits just left of the sample's.

Read off a and b

Why: the lightest and heaviest possible totals.

\[ a = 0, \quad b = 14 \]

Write the distribution

Why: X is the total weight of fish on one boat, in pounds.

\[ X \sim U(0, 14) \]

The mean is the midpoint

Why: average the endpoints.

\[ \mu = \frac{0 + 14}{2} = 7 \text{ pounds} \]

The standard deviation

Why: width over root twelve.

\[ \sigma = \frac{14 - 0}{\sqrt{12}} \approx 4.04 \text{ pounds} \]

Recomputed from Table 5.2, the sample mean is 7.857 pounds, which the book rounds to 7.9, and the sample standard deviation is 4.326, which checks. The sample sits a little to the right of the model's centre and is a little more spread out: the dot plot has heavy stacks at 10 and 11 and a gap at 8.

With only 35 boats, differences of this size are what samples do. The model's claim is only that no weight between 0 and 14 is favoured, and the dot plot shows values across the whole range.

Check against the data

Why: the model's mean and spread should be near the sample's.

\[ 7 \text{ vs } 7.86, \qquad 4.04 \text{ vs } 4.33 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 296 — Try It 5.2 and Table 5.2

31. Example 5.4(b): the average wait for the bus

Worked example

For the bus, X ~ U(0, 15) minutes. On average, how long must a person wait? Find the mean and the standard deviation.

The mean

Why: the midpoint of 0 and 15.

\[ \mu = \frac{0 + 15}{2} = 7.5 \text{ minutes} \]

The variance

Why: the width squared over twelve.

\[ \sigma^{2} = \frac{(15 - 0)^{2}}{12} = \frac{225}{12} = 18.75 \]

The standard deviation

Why: the square root of the variance.

\[ \sigma = \sqrt{18.75} \approx 4.33 \text{ minutes} \]

The book prints 4.3; to two decimal places it is 4.33. In context: on average a person waits 7.5 minutes, and a typical wait differs from that average by about four and a third minutes.

That spread is large relative to the mean, and it should be. Waits anywhere from 0 to 15 minutes are equally likely, so a wait of 1 minute and a wait of 14 are both ordinary. Compare the Poisson counts of Chapter 4, where the spread was only the square root of the mean.

It also says something practical. If you plan to arrive at your stop and catch a bus within the average wait, you will miss that target about half the time, because the mean is also the median here. A plan based on the average alone is a coin toss.

Check with the one-standard-deviation band

Why: from 7.5 minus 4.33 to 7.5 plus 4.33 should hold about 58 percent of the width.

\[ \frac{11.83 - 3.17}{15} = \frac{8.66}{15} \approx 0.577 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 298 — Example 5.4(b); the book rounds σ to 4.3

32. Trap: stopping before the square root

Trap

The trap

For the bus, X ~ U(0, 15), a first attempt at the standard deviation writes:

\[ \sigma = \frac{(15 - 0)^{2}}{12} \]

\[ \sigma = 18.75 \text{ minutes} \]

A second attempt forgets the square as well as the root:

\[ \sigma = \frac{15}{12} = 1.25 \]

Wrong both times.

The fix

The first number is the variance, in minutes squared. The second divides by 12 instead of its square root and is far too small.

\[ \sigma = \sqrt{18.75} \]

\[ \quad \approx 4.33 \text{ minutes} \]

A spread check catches both. Every wait lies between 0 and 15, so no wait can be more than 7.5 minutes from the mean of 7.5; a standard deviation of 18.75 is impossible. And a spread of 1.25 would mean most waits sit near 7.5, which a flat distribution does not do. The standard deviation of a uniform distribution is always about 29 percent of the width.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 298 — Example 5.4(b)

33. Move the endpoints and watch the rectangle

Tweak it

Drag a and b. The rectangle keeps an area of one, so its height responds to the width. The readout shows the height, the mean and the standard deviation for the interval you have set.

Parameter explorer

Set a to 0 and b to 23 to see the smiling times. Then slide a up by 5 and b up by 5, keeping the width at 23. Which of the three readouts change, and which stay the same? Finally, squeeze the interval to its narrowest, a at 10 and b at 14, and watch the height and the spread.

\[ f(x) = \frac{1}{{b} - {a}}, \quad \mu = \frac{{a} + {b}}{2}, \quad \sigma = \frac{{b} - {a}}{\sqrt{12}} \]

  • a — from 0 to 10: a, lowest value
  • b — from 14 to 30: b, highest value

Sliding both endpoints by the same amount moves the rectangle without changing its shape. The height and the standard deviation stay put, because they depend only on the width; the mean moves by exactly the amount you slid. That is the algebra of the last few slides made visible.

Now narrow the interval as far as the sliders allow, a at 10 and b at 14. The rectangle shoots up to a height of one quarter and the standard deviation falls to about 1.15. A narrow uniform distribution is tall and concentrated; a wide one is low and spread out, and the area never changes.

Notice that the height and the standard deviation always move in opposite directions. Their product is one over the square root of 12, about 0.29, for every uniform distribution.

34. Shifting the interval by 100

Prediction

Two machines fill bags. On the first, the fill error is uniform between 0 and 10 grams. On the second, the fill weight is uniform between 100 and 110 grams. Predict how their means and standard deviations compare.

Predict first

How do U(0, 10) and U(100, 110) compare?

  • different means, same standard deviation
  • same mean, different standard deviations
  • different means and different standard deviations
  • the second has a standard deviation ten times larger

Correct: different means, same standard deviation

Why: Both intervals are 10 grams wide, so both have standard deviation 10 over the square root of 12, about 2.89 grams. Only the mean moves: 5 grams for the first and 105 grams for the second. Shifting a distribution moves its centre and leaves its spread alone.

\[ \sigma = \frac{10}{\sqrt{12}} \approx 2.89 \text{ for both}, \qquad \mu = 5 \text{ and } 105 \]

If you picked the option with a spread ten times larger, you were probably thinking of the second machine as bigger in every way. The formula only sees the width. It would be ten times larger if the second interval ran from 0 to 100, which is a different situation altogether.

The same idea explains why the baseball season, uniform between 447 and 521 hours, has a standard deviation of 21.36 hours, not something in the hundreds. The width is 74 hours; the 447 only says where the interval starts.

35. Percentiles: solving for the boundary

Section

Part 4

36. Given the area, find the boundary

Concept

So far you have been handed a boundary and asked for an area. A percentile question runs the other way: it hands you an area and asks where the boundary is. The kth percentile is the value with that proportion of the area to its left.

Call the unknown boundary k. The area to its left is a rectangle with base from a to k, so you write that area and set it equal to the proportion you were given.

\[ P(x < k) = (k - a)\left(\frac{1}{b - a}\right) = p \]

Multiply both sides by the width

Why: clear the fraction.

\[ k - a = p(b - a) \]

Add a to both sides

Why: k on its own.

\[ k = a + p(b - a) \]

Read the result as a walk: start at a, then go the proportion p of the way along the interval. The 90th percentile is nine tenths of the way from a to b; the 50th is halfway, which is the mean you found in Part 3.

You do not need to memorise the last line. The book always writes the area equation and solves it, because that works for a left-hand area, a right-hand area, or anything else you can shade. What matters is that the base starts at a, not at 0, and that the equation says area equals proportion.

The book adds that k is sometimes called a critical value, a name you will meet again in the chapters on hypothesis testing, where a boundary cutting off a given area is exactly what a test needs.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, pp. 297, 299 — Example 5.3(b) and Example 5.4(c): k is sometimes called a critical value

37. Example 5.3(b): the 90th percentile of smiling times

Worked example

Find the 90th percentile for an eight-week-old baby's smiling time, with X ~ U(0, 23). Ninety percent of smiling times fall below the 90th percentile, k.

Write the area statement

Why: ninety percent of the area lies to the left of k.

\[ P(x < k) = 0.90 \]

Write the area as base times height

Why: the base runs from 0 to k.

\[ (k - 0)\left(\frac{1}{23}\right) = 0.90 \]

Multiply both sides by 23

Why: clear the fraction.

\[ k = 0.90 \times 23 \]

Simplify

Why: nine tenths of 23.

\[ k = 20.7 \text{ seconds} \]

Figure (svg): The uniform density of height 1 over 23 from 0 to 23. The region from 0 to 20.7 is shaded and labelled area 0.90. A vertical line at 20.7 is labelled k. The unshaded strip from 20.7 to 23 is labelled 0.10.

The unknown is now the boundary, not the area. Slide k to the right until the shaded part holds exactly nine tenths of the rectangle.

In context: 90 percent of smiling times of eight-week-old babies are shorter than 20.7 seconds, and 10 percent are longer. In the actual sample, 51 of the 55 times, about 93 percent, lie below 20.7, close to the model's 90.

Notice the difference from Part 2 in what is unknown. Here the answer is a smiling time in seconds, not a probability, so it must lie between 0 and 23. An answer like 0.9 or 0.039 is a signal that the roles of area and boundary were swapped.

Check by computing the area back

Why: the rectangle from 0 to 20.7 must hold 0.90.

\[ (20.7)\left(\frac{1}{23}\right) = 0.90 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 297 — Example 5.3(b)

38. A percentile reads the ramp backwards

Intuition

Part 2 drew the cumulative ramp: go up from a value x to the ramp, then across to read the area to its left. A percentile question starts on the other axis.

Figure (svg): The cumulative graph for the smiling times, a straight ramp from 0 at x equals 0 to 1 at x equals 23. An arrow starts on the vertical axis at 0.9, runs across to the ramp, and then an arrow runs down to the x-axis at 20.7.

A probability question goes up from x to the ramp and across. A percentile question goes the other way: across from the area, then down to x.

Start at 0.9 on the vertical axis, go across until you meet the ramp, then drop down to the horizontal axis. You land at 20.7 seconds. The algebra on the last slide, k equals 0.9 times 23, is exactly this trip, because the ramp is a straight line.

\[ \frac{k - a}{b - a} = p \quad \Longleftrightarrow \quad k = a + p(b - a) \]

The picture also shows why every percentile of a uniform distribution is evenly spaced. The 10th, 20th, 30th and so on up to the 90th percentile of the smiling times are 2.3 seconds apart, because equal steps up the vertical axis meet a straight ramp at equal steps along the horizontal one.

For the curved densities of later chapters the ramp bends, the spacing is uneven, and you will need a table or software to make the trip. The idea, across from the area and down to the value, never changes.

39. Example 5.4(c): the 90th percentile of the bus wait

Worked example

For the bus, X ~ U(0, 15) minutes. Ninety percent of the time, the time a person must wait falls below what value?

Recognise the question

Why: ninety percent below a value is the 90th percentile; let k be that value.

\[ P(x < k) = 0.90 \]

Write the area

Why: base from 0 to k, height one fifteenth.

\[ (k - 0)\left(\frac{1}{15}\right) = 0.90 \]

Multiply both sides by 15

Why: clear the fraction.

\[ k = 0.90 \times 15 \]

Simplify

Why: nine tenths of 15.

\[ k = 13.5 \text{ minutes} \]

Figure (svg): The uniform density of height 1 over 15 from 0 to 15 minutes. The region from 0 to 13.5 is shaded and labelled 0.90. A vertical line at 13.5 is labelled k equals 13.5.

Nine tenths of the fifteen-minute rectangle ends at 13.5 minutes. Only one wait in ten runs longer.

The book's interpretation is worth copying word for word: ninety percent of the time, a person must wait at most 13.5 minutes. The words at most carry the meaning of percentile, and minutes gives the units.

This is the number a transit planner uses. A traveller who allows 13.5 minutes for the wait will be late for their connection only one time in ten; allowing only the mean, 7.5 minutes, would fail half the time.

Check with the area to the right

Why: the unshaded strip from 13.5 to 15 must hold the other ten percent.

\[ (15 - 13.5)\left(\tfrac{1}{15}\right) = \frac{1.5}{15} = 0.10 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 299 — Example 5.4(c)

40. Example 5.6(c): the 30th percentile of furnace repairs

Worked example

For the furnace, X ~ U(1.5, 4) hours with height 0.4. Find the 30th percentile of repair times.

Write the area statement

Why: thirty percent of the area lies to the left of k.

\[ P(x < k) = 0.30 \]

Write the area as base times height

Why: the base starts where the rectangle starts, at 1.5.

\[ (k - 1.5)(0.4) = 0.30 \]

Divide both sides by 0.4

Why: undo the height.

\[ k - 1.5 = 0.75 \]

Add 1.5 to both sides

Why: undo the start of the interval.

\[ k = 2.25 \text{ hours} \]

Figure (svg): The uniform density of height 0.4 from 1.5 to 4 hours. The region from 1.5 to 2.25 is shaded, labelled 0.30, with a vertical line at k equals 2.25. The base of the shaded piece is labelled k minus 1.5.

The base of the shaded piece runs from 1.5, where the rectangle starts, to k. That is why the equation reads k minus 1.5, not k.

In context, as the book puts it: 30 percent of repair times are 2.25 hours or less. A student who writes k times 0.4 equals 0.3 gets k equal to 0.75 hours, which is shorter than any repair can take, so it cannot be a percentile of this distribution.

Notice the order of the two solving moves. You divide by the height first, which turns the area 0.3 into a width, 0.75 hours; then you add the start of the interval to turn that width into a position on the axis. Width first, then position, every time.

Check with the walk-along form

Why: start at 1.5 and go three tenths of the 2.5-hour width.

\[ 1.5 + 0.30(2.5) = 1.5 + 0.75 = 2.25 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 301 — Example 5.6(c), Figure 5.20

41. Example 5.6(d): the longest 25 percent of repairs

Worked example

The longest 25 percent of furnace repair times take at least how long? In other words, find the minimum time for the longest 25 percent of repairs. What percentile does this represent?

Write the area statement

Why: the longest quarter is a right-hand area: a quarter of repairs take longer than k.

\[ P(x > k) = 0.25 \]

Write the area as base times height

Why: the base runs from k up to the end of the interval, 4.

\[ (4 - k)(0.4) = 0.25 \]

Divide both sides by 0.4

Why: undo the height.

\[ 4 - k = 0.625 \]

Subtract 4 from both sides

Why: the book's move.

\[ -k = -3.375 \]

Multiply by negative one

Why: k on its own.

\[ k = 3.375 \text{ hours} \]

Figure (svg): The uniform density of height 0.4 from 1.5 to 4 hours. The region from 3.375 to 4 on the right is shaded and labelled 0.25, the longest repairs. The unshaded region from 1.5 to 3.375 is labelled 0.75.

The area named in the question is on the right. The same boundary has three quarters of the area on its left, so it is the 75th percentile.

In context: the longest 25 percent of furnace repairs take at least 3.375 hours. Since a quarter of repairs lie above 3.375, three quarters lie below it, so 3.375 hours is the 75th percentile. The question named the area on the right; percentiles always count the area on the left.

You could also skip the right-hand equation entirely: a quarter on the right means three quarters on the left, so solve for the 75th percentile directly. Both routes are fine. What is not fine is writing 0.25 into a left-hand equation, which the next slide shows.

Check with the left-hand area

Why: the base from 1.5 to 3.375 must hold the other three quarters.

\[ (3.375 - 1.5)(0.4) = (1.875)(0.4) = 0.75 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, pp. 301-302 — Example 5.6(d), Figure 5.21

42. Trap: the longest 25 percent is not the 25th percentile

Trap

The trap

For the longest quarter of furnace repairs, a first attempt shades the left of k:

\[ (k - 1.5)(0.4) = 0.25 \]

\[ k - 1.5 = 0.625 \]

\[ k = 2.125 \text{ hours} \]

and calls it the time the longest 25 percent take at least.

Wrong. That is the 25th percentile, the boundary of the shortest quarter.

The fix

Longest means the right-hand end of the rectangle. Put the area where the words put it, then solve.

\[ (4 - k)(0.4) = 0.25 \]

\[ k = 3.375 \text{ hours} \]

The wrong answer says three quarters of repairs take at least 2.125 hours, which is true but answers a different question. Before writing the equation, sketch the rectangle and shade the quarter the question describes: longest, slowest and more than shade the right; shortest, fastest and less than shade the left. Then, if asked for the percentile, count the area to the left of your k.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 302 — Example 5.6(d)

43. Try It 5.4(e): the 65th percentile of a baseball season

Fill the middle

Return to the baseball season, X ~ U(447, 521) hours, 74 hours wide. Find the 65th percentile by walking 65 percent of the way along the interval from its start.

Fill in the blanks

k = 447 + 0.65(74) = 495.1

Why: The width is 521 minus 447, which is 74 hours, and 65 percent of 74 is 48.1. Starting from 447, that reaches 495.1 hours. Written as an area equation it is k minus 447, times one over 74, equal to 0.65.

In context: 65 percent of teams' seasons total 495.1 hours of games or fewer. The answer sits between 447 and 521, as a percentile must, and a little past the middle, 484, as a percentile above the 50th should.

The common slip here is 0.65 times 521, which gives 338.65, an impossible season far below the lowest value of 447. That is the percentile version of measuring the base from 0. The other half of Try It 5.4, part (c), is the mean, 484 hours, and the standard deviation, 74 over the square root of 12, about 21.36 hours.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 299 — Try It 5.4, parts (c) and (e)

44. Push the percentile to its limits

Edge cases

The walk-along form, k equals a plus p times the width, works for any proportion p. Test it at the edges, where a formula either makes sense or tells you that you asked a meaningless question.

Discussion prompt

For the furnace, uniform between 1.5 and 4 hours, find the 0th, 50th and 100th percentiles. Then try the formula with p equal to 1.2, and say what goes wrong. Finally, compute the probability that a repair takes less than 5 hours with base times height, and fix the answer.

Both failures have the same cause: the formula only knows about the rectangle, and you have walked off it. A percentile must lie between a and b, and a probability between 0 and 1. If a result breaks either rule, the fix is to cut back to the rectangle, not to report the number.

The 50th percentile is worth remembering on its own. For a uniform distribution the median and the mean are the same number, the midpoint. That symmetry fails for most of the distributions that follow, so it is a useful habit to check which of the two a question is asking for.

45. Conditional probability: reducing the sample space

Section

Part 5

46. What you already know throws part of the interval away

Concept

Example 5.3(c) asks for the probability that a baby smiles more than 12 seconds, knowing that the baby has already smiled more than 8 seconds. This is a conditional probability, and the word knowing changes the problem.

\[ P(x > 12 \mid x > 8) \]

Read the vertical bar as given that. Once you know the smile has passed 8 seconds, every smiling time below 8 is no longer possible for this baby. The sample space shrinks from 0 to 23 down to 8 to 23.

Within what is left, is any time more likely than another? No. The original model said every time from 0 to 23 was equally likely, so every time from 8 to 23 still is. The new sample space is again uniform, just on a shorter interval.

\[ X \mid (x > 8) \sim U(8, 23) \]

\[ f(x) = \frac{1}{23 - 8} = \frac{1}{15} \quad \text{for } 8 < x < 23 \]

This is the first of the book's two ways to solve the problem: redraw the rectangle on the reduced interval, with the height that the new width forces, and then answer an ordinary base-times-height question on it. The second way, next, keeps the original rectangle and uses the conditional formula from Chapter 3.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 297 — Example 5.3(c), first way

47. Conditioning redraws the rectangle

Picture it

Here are both rectangles on one set of axes: the original, dashed, and the reduced one, solid. Watch what happens to the height.

Figure (svg): One graph with two rectangles. A dashed outline shows the original density, height 1 over 23 from 0 to 23. The region from 0 to 8 is washed red and labelled ruled out. A solid, taller rectangle of height 1 over 15 runs from 8 to 23, labelled new sample space. An arrow points up from the old height to the new one.

Knowing the smile lasted more than 8 seconds throws away the first 8 seconds. The 15 seconds that remain must still carry all the probability, so the rectangle rises to one fifteenth.

The red wash marks the first 8 seconds, ruled out by what you know. The 15 seconds that remain must still carry a total probability of one, because the smile certainly lasts some time between 8 and 23. A rectangle 15 wide with area one must be one fifteenth tall.

\[ \frac{1}{23} \approx 0.0435 \quad \longrightarrow \quad \frac{1}{15} \approx 0.0667 \]

So information you are given does not just remove area; it redistributes the whole probability over what is left. Every remaining value becomes more likely, in the same proportion, by a factor of 23 over 15.

That factor is the key to seeing why the two methods agree. Keeping the old height and forgetting to raise it is the error in the trap later in this part: the answer would come out too small by exactly that factor.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 297 — Figure 5.13, the new sample space

48. Example 5.3(c), first way: reduce the sample space

Worked example

Find the probability that a random eight-week-old baby smiles more than 12 seconds, knowing that the baby smiles more than 8 seconds.

Write the conditional

Why: the event, then the bar, then what is known.

\[ P(x > 12 \mid x > 8) \]

Reduce the sample space

Why: the smile is known to lie between 8 and 23.

\[ X \mid (x > 8) \sim U(8, 23) \]

Write the new height

Why: one over the new width.

\[ f(x) = \frac{1}{23 - 8} = \frac{1}{15} \]

Base times new height

Why: from 12 up to 23.

\[ P(x > 12 \mid x > 8) = (23 - 12)\left(\frac{1}{15}\right) \]

Simplify

Why: eleven fifteenths.

\[ = \frac{11}{15} \approx 0.7333 \]

Figure (svg): The reduced density: a rectangle of height 1 over 15 from 8 to 23 seconds. The region from 12 to 23 is shaded and labelled base 11, area 11 over 15, about 0.7333. The region from 8 to 12 is unshaded.

On the new rectangle the question is an ordinary one: base 11, height one fifteenth.

In context: of babies whose smile has already lasted more than 8 seconds, about 73 percent will smile for more than 12 seconds in total. Compare that with the unconditional chance of a smile longer than 12 seconds, 11 over 23, about 0.48. Knowing the smile is already long makes a longer one more likely.

The whole method fits in one sentence: throw away what you know did not happen, spread the probability evenly over what is left, and measure the event on that new rectangle. The only new arithmetic is the height, and the width that forces it is the width of what you know.

Check with the leftover piece

Why: on the new rectangle, 8 to 12 must hold the rest.

\[ (12 - 8)\left(\tfrac{1}{15}\right) = \frac{4}{15}, \quad \frac{11}{15} + \frac{4}{15} = 1 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 297 — Example 5.3(c), first way, Figure 5.13

49. Reading the conditional formula

Notation

The second way uses the conditional probability formula from Chapter 3 on the original distribution. Decode it once, with the smiles in mind.

Annotate

On: \( P(A \mid B) = \frac{P(A \text{ AND } B)}{P(B)} \)

  • The event you want: here, the smile lasts more than 12 seconds.
  • What you know: here, the smile lasts more than 8 seconds. It becomes the new whole.
  • Both at once. A smile longer than 12 seconds is automatically longer than 8, so here A AND B is just x greater than 12.
  • Rescales so that the known event carries total probability one. This division is exactly the taller rectangle of the first way.

The step people skip is simplifying A AND B. When the event you want sits inside what you know, as more than 12 sits inside more than 8, the AND leaves just A. That is not true in general, so decide it each time with a sketch of the two shaded regions.

\[ (x > 12) \text{ AND } (x > 8) = (x > 12) \]

Compare the counterexample later in this part, where the event is x less than 12 and the known fact is again x greater than 8. There the AND is the strip from 8 to 12, which is neither event on its own. Sketch both regions and shade the overlap; the overlap is the numerator.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §3.2 formula, used in §5.2, p. 297 — P(A|B) = P(A AND B)/P(B)

50. Example 5.3(c), second way: the conditional formula

Worked example

Solve the same problem on the original distribution, X ~ U(0, 23), with A the event x greater than 12 and B the event x greater than 8.

Name A and B

Why: the event and what is known.

\[ A: x > 12, \qquad B: x > 8 \]

Simplify A AND B

Why: every x above 12 is above 8.

\[ A \text{ AND } B: x > 12 \]

Area of A AND B

Why: base from 12 to 23 on the original rectangle.

\[ P(x > 12) = (23 - 12)\left(\tfrac{1}{23}\right) = \frac{11}{23} \]

Area of B

Why: base from 8 to 23.

\[ P(x > 8) = (23 - 8)\left(\tfrac{1}{23}\right) = \frac{15}{23} \]

Divide

Why: the conditional formula.

\[ P(x > 12 \mid x > 8) = \frac{11/23}{15/23} \]

Cancel the twenty-thirds

Why: the old height drops out.

\[ = \frac{11}{15} \approx 0.7333 \]

Figure (svg): The original density of height 1 over 23 from 0 to 23. The region from 8 to 23 is lightly shaded and labelled B: x greater than 8, area 15 over 23. Inside it, the region from 12 to 23 is darker and labelled A and B: x greater than 12, area 11 over 23.

The darker piece sits inside the lighter one. The conditional probability is the dark area as a share of the whole shaded area.

The picture is the book's Figure 5.14: the dark region is A AND B, the whole shaded region is B, and the answer is dark as a share of shaded. The twenty-thirds cancel because both areas have the same height, so only the bases matter, 11 over 15.

This way needs no new density, which is its advantage: you never redraw anything, you only shade twice on the rectangle you already have. Its cost is one extra step, simplifying the AND, which is where it goes wrong when the event is not inside the known region.

Check against the first way

Why: both methods must give the same value.

\[ \frac{11}{15} = 0.7333 \quad \text{both ways} \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, pp. 297-298 — Example 5.3(c), second way, Figure 5.14

51. Why the two ways always agree

Intuition

The two methods look different, but for a uniform distribution they are the same arithmetic in a different order. Write both in general, for the event x greater than d given x greater than c, with c less than d, inside U(a, b).

The first way redraws the rectangle from c to b and takes the base from d to b.

\[ \text{first way: } (b - d)\left(\frac{1}{b - c}\right) = \frac{b - d}{b - c} \]

The second way takes two areas on the original rectangle and divides them. The original height, one over b minus a, appears in both and cancels.

\[ \text{second way: } \frac{(b - d)/(b - a)}{(b - c)/(b - a)} = \frac{b - d}{b - c} \]

So a conditional probability for a uniform distribution is always a ratio of two widths: the width of what you want, over the width of what you know. The original a has disappeared from the answer entirely, which makes sense: once you know x is above c, where the interval used to start no longer matters.

Use whichever way you find clearer, but the book recommends doing it both ways at first, because each checks the other. The first way is quicker to draw; the second way is the one that carries over to the curved distributions of later sections, where no redrawn rectangle exists.

52. Trap: keeping the original height after conditioning

Trap

The trap

For the smiling times, a first attempt at the probability of more than 12 seconds given more than 8 writes:

\[ (23 - 12)\left(\frac{1}{23}\right) \]

\[ = \frac{11}{23} \approx 0.4783 \]

Wrong. That ignores the condition entirely.

The fix

That is the unconditional probability of a smile longer than 12 seconds. Knowing the smile has passed 8 seconds shrinks the sample space, and the height has to rise to match.

\[ (23 - 12)\left(\frac{1}{15}\right) \]

\[ = \frac{11}{15} \approx 0.7333 \]

A quick test catches it: compute the probability of the known event itself on your rectangle. Given x greater than 8, the probability that x is greater than 8 must be one. With the old height it comes to 15 over 23, which shows the rectangle was never redrawn.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 297 — Example 5.3(c)

53. Example 5.5: the donut, both parts

Worked example

The time a nine-year-old takes to eat a donut is between 0.5 and 4 minutes, X ~ U(0.5, 4). Find (a) the probability that a child takes more than two minutes, and (b) the probability that a child takes more than two minutes given that the child has already been eating for more than 1.5 minutes.

Part (a): the height

Why: the width is 3.5 minutes.

\[ f(x) = \frac{1}{4 - 0.5} = \frac{1}{3.5} \]

Part (a): base times height

Why: from 2 up to 4.

\[ P(x > 2) = (4 - 2)\left(\frac{1}{3.5}\right) \approx 0.5714 \]

The book words part (a) as eats a donut in at least two minutes, meaning takes at least two minutes; its answer, 0.5714, is the probability of more than two. The endpoint makes no difference for a continuous variable.

Part (b): reduce the sample space

Why: you are no longer starting at 0.5; the starting point is 1.5.

\[ X \mid (x > 1.5) \sim U(1.5, 4), \quad f(x) = \frac{1}{2.5} = 0.4 \]

Part (b): base times new height

Why: from 2 up to 4.

\[ P(x > 2 \mid x > 1.5) = (4 - 2)(0.4) = 0.8 \]

Figure (svg): Two graphs. Left: the original donut density, height 1 over 3.5 from 0.5 to 4 minutes, with the region from 2 to 4 shaded, area about 0.5714. Right: the reduced density, height 1 over 2.5, that is 0.4, from 1.5 to 4, with the same region from 2 to 4 shaded, area 0.8.

Same shaded base, 2 minutes, both times. What changes is the height: once the first minute and a half is known to be past, the rectangle is narrower and so taller.

In context: about 57 percent of children take more than two minutes to eat a donut, but among children still eating after a minute and a half, 80 percent will take more than two minutes in total.

Check part (b) with the conditional formula

Why: the book's second way, on the original rectangle.

\[ \frac{P(x > 2)}{P(x > 1.5)} = \frac{2/3.5}{2.5/3.5} = \frac{2}{2.5} = 0.8 \]

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, pp. 299-300 — Example 5.5, Figure 5.17

54. Which way does the condition push?

Prediction

For the smiling times, compare two probabilities without computing: a smile longer than 12 seconds, and a smile longer than 12 seconds given that it has already lasted 8.

Predict first

How does the probability of a smile longer than 12 seconds change once you know the smile has lasted more than 8 seconds?

  • it goes up
  • it goes down
  • it stays the same
  • it cannot be decided without the data

Correct: it goes up

Why: Without the condition, the chance is 11 out of 23 seconds of width, about 0.48. Knowing the smile is past 8 seconds removes 8 seconds of width that were all short smiles, and leaves 11 out of 15, about 0.73. Throwing away values that are not in the event raises the event's share.

\[ \frac{11}{23} \approx 0.48 \quad \longrightarrow \quad \frac{11}{15} \approx 0.73 \]

The condition removed only values that were outside the event: every smile shorter than 8 seconds was also shorter than 12. When you throw away only failures, the successes become a bigger share of what is left.

Put numbers on it. Of every 23 seconds of possible smiling time, 11 belong to the event. The condition removes 8 seconds, none of which belonged to the event, so the event keeps its 11 out of the 15 that remain.

Keep that reasoning, not the conclusion. The next slides show a condition that pushes a probability down, and the difference is entirely in which values the condition throws away.

55. Try It 5.5: the quiz, before any arithmetic

Step zero

The time a student takes to finish a quiz is uniform between 6 and 15 minutes, X ~ U(6, 15). The book asks two questions: the probability that a student needs at least 8 minutes, and the probability that a different student needs at least 8 minutes given that she has already taken more than 7.

Figure (svg): The uniform density of height 1 over 9 from 6 to 15 minutes. Dashed vertical lines mark 7, where the condition starts, and 8, where the event starts.

Two boundaries, two jobs. The line at 8 is where the event starts; the line at 7 is where the information you are given starts.

Discussion prompt

Before computing, say what the rectangle is for each question: its interval and its height. Then compute both probabilities.

The step zero is deciding which rectangle you are on. The two questions shade the same 7-minute base, and only the height differs. Getting the rectangle right is the whole problem; the multiplication afterwards takes a second.

As a check by the second way: seven ninths divided by eight ninths is seven eighths. The ninths cancel, as the old height always does.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 300 — Try It 5.5

56. Does knowing more always raise the probability?

Counterexample

After the smiles and the donut, a classmate says: knowing that x is greater than some value always makes a probability bigger. Test that claim with the smiling times.

Discussion prompt

Find a probability about the smiling times, X ~ U(0, 23), that goes down when you learn that the smile lasted more than 8 seconds. Compute both values.

The claim fails because a condition can remove values from either side. In the smiling example of Part 5 it removed only values outside the event, so the event's share rose. Here it removes values inside the event, so the share falls. A single counterexample is enough to sink a claim that says always.

What is always true is the recipe: redraw the rectangle on what is known, and take the event's share of it. Whether the answer rises or falls is something the sketch tells you, not something to assume.

57. Putting it together

Section

Part 6

58. One rectangle, five answers

Picture it

Example 5.6 asks five questions about one furnace distribution, X ~ U(1.5, 4). Here are all the answers on a single rectangle.

Figure (svg): The furnace density, height 0.4 from 1.5 to 4 hours. Vertical lines mark the 30th percentile at 2.25, the mean at 2.75 and the 75th percentile at 3.375. A double-headed arrow shows one standard deviation, 0.72, on each side of the mean.

Every question in the section is a statement about this one rectangle: an area, a centre, a spread, or a boundary that cuts off a given area.

Two of the answers are areas, 0.8 for more than two hours and 0.6 for less than three. Two are boundaries, the 30th percentile at 2.25 hours and the 75th at 3.375. And two describe the whole rectangle: the mean at its midpoint, 2.75 hours, and the standard deviation, about 0.72 hours, drawn as the arrows either side of the mean.

\[ \mu = \frac{1.5 + 4}{2} = 2.75, \qquad \sigma = \frac{4 - 1.5}{\sqrt{12}} \approx 0.7217 \]

Notice how the percentiles and the mean line up along the interval. The 30th percentile is three tenths of the way along, the mean is halfway, and the 75th percentile is three quarters of the way. Evenly spaced proportions give evenly spaced boundaries, because the density is flat.

Before the arithmetic of any question about a uniform distribution, place the answer roughly on this picture. It takes a few seconds and catches nearly every setup error in the section.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, pp. 300-302 — Example 5.6(a) to (e); the values for part (e) are computed here

59. One new rectangle, six questions

Comparison

A delivery is equally likely to arrive any time between 2 and 10 days after an order, X ~ U(2, 10). Fill in each value. Every row is the same rectangle, 8 days wide.

Comparison matrix

questionset-upvalue
height of the density1/(10 − 2)0.125
P(x < 4)(4 − 2)(1/8)0.25
P(x > 7)(10 − 7)(1/8)0.375
40th percentile2 + 0.40(8)5.2
mean(2 + 10)/26
P(x > 7 | x > 4)(10 − 7)(1/6)0.5

Read the set-up column before filling in the values: every base starts at 2 or ends at 10, never at 0, and the last row's height is one sixth, not one eighth, because the known fact that the delivery is past day 4 leaves a 6-day rectangle.

The standard deviation is the one question not in the table: 8 over the square root of 12, about 2.31 days. Notice also that the 40th percentile, 5.2 days, sits just before the mean of 6, as a percentile below the 50th must.

Each row took one line of arithmetic because each question was first turned into a statement about the rectangle. That translation is the skill the section teaches.

60. Match each question to its rectangle calculation

Matching

For X ~ U(0, 20), match each question to the calculation that answers it. Several answers share numbers, so read the set-ups, not just the results.

Match the pairs

  • a. P(x < 5)
  • b. P(x > 5)
  • c. the 25th percentile
  • d. the mean
  • e. P(x > 15 | x > 5)
  • v. (5)(1/20) = 0.25
  • w. (15)(1/20) = 0.75
  • x. k(1/20) = 0.25, so k = 5
  • y. (0 + 20)/2 = 10
  • z. (5)(1/15) ≈ 0.333

Why: The probability of less than 5 is a base of 5 times one twentieth. More than 5 is the other 15 of the width. The 25th percentile solves an area equation for the boundary, and lands at 5 because 5 is a quarter of the way along. The mean is the midpoint. The conditional redraws the rectangle from 5 to 20 with height one fifteenth.

Two pairs are mirror images. The probability that x is less than 5 is 0.25, and the 25th percentile is 5: one question gives the boundary and asks for the area, the other gives the area and asks for the boundary. On the cumulative ramp they are the same point, read in opposite directions.

The conditional row is the one to slow down on. The base from 15 to 20 is 5 wide, the same as the base for less than 5, but the height is one fifteenth because the known fact removed the first 5 units of the interval.

61. Pattern: solving any uniform problem

Pattern

Every problem in this section ran through the same short sequence. The branch in the middle depends only on what the question asks for: an area, a boundary, or an area after you have been told something. Decide which before you write a single number, because the three branches write different equations from the same rectangle.

Figure (svg): A flow diagram. Name X, a and b; then draw the rectangle with height one over b minus a. From the rectangle, three branches: an area question uses base times height; a percentile question sets base times height equal to the proportion and solves for k; a given-that question redraws the rectangle on the reduced interval. All three end in a box: interpret in context.

Two moves at the top are always the same. The branch depends on what the question wants: an area, a boundary, or an area after something is already known.
  1. Name X, a and b. Define X in words with units, and write X ~ U(a, b), lowest value first.
  2. Draw the rectangle from a to b with height one over b minus a. Mark a and b on the axis.
  3. An area? Shade it, clip the base to the rectangle, and multiply base by height.
  4. A boundary? Shade the area the words describe, write base times height equal to the proportion, and solve for k.
  5. Given that? Redraw the rectangle on what is known, with its new height, then find the area.
  6. Mean and spread: the midpoint, and the width over the square root of 12.
  7. Interpret in context, in the units of X.

\[ f(x) = \frac{1}{b - a}, \quad P = (\text{base})(\text{height}), \quad \mu = \frac{a + b}{2}, \quad \sigma = \frac{b - a}{\sqrt{12}} \]

Every step has a free check: the whole rectangle has area one; a probability lies between 0 and 1; a percentile lies between a and b; a conditional recomputed by the formula agrees with the redrawn rectangle. Use them. The arithmetic is easy enough that nearly every wrong answer in this section is a setup error that one of these checks would catch.

62. The bus you do not have a timetable for

Real world

Uniform distributions show up whenever a random moment meets a fixed cycle: arriving at a stop without checking the schedule, glancing at a traffic light, starting to listen to a looping announcement. The book's bus example is the everyday version.

Discussion prompt

A bus comes exactly every 12 minutes and you arrive at a random moment. Write the distribution of your wait, find your average wait and the chance of waiting more than 10 minutes. Then explain why, if the gaps between buses themselves vary but still average 12 minutes, your real average wait is longer than your answer.

The second part is the transfer prompt from this section's homework, and it is a genuinely useful piece of reasoning. The uniform model is exactly right for a fixed cycle, and it tells you the best case. Real bus systems that bunch up do worse than the model, which is why transit agencies publish not only how often buses come but how evenly.

Notice which assumption carried the answer: that your arrival time is unrelated to the timetable. If you check the schedule and arrive two minutes early, your wait is not uniform at all, and the model no longer applies.

63. Check: the furnace's mean and standard deviation

Check

Example 5.6(e) asks for the mean and standard deviation of the furnace repair times. The values are computed here, since the book's are lost from the extracted text.

Check your understanding

For repair times uniform between 1.5 and 4 hours, what are the mean and the standard deviation?

  • A. μ = 2.75, σ ≈ 0.72 (correct)
  • B. μ = 2.75, σ ≈ 0.52
  • C. μ = 1.25, σ ≈ 0.72
  • D. μ = 2, σ ≈ 1.15

Answer: A

Why: The mean is the midpoint, 1.5 plus 4 over 2, which is 2.75 hours. The width is 2.5 hours, and the standard deviation is the width over the square root of twelve, about 0.7217 hours.

Why B tempts people
That is the variance, 2.5 squared over 12, about 0.52 hours squared. Take its square root to get the standard deviation.
Why C tempts people
That halves the width instead of averaging the endpoints. The mean must lie inside the interval, and 1.25 is below its start at 1.5.
Why D tempts people
That ignores a: 4 over 2 for the mean, and 4 over the square root of 12 for the spread. Both use b alone instead of the interval from 1.5 to 4.

The mean has a built-in check: it must lie inside the interval, and in the middle of it. That rules out 1.25 and 2 at once, since the midpoint of 1.5 and 4 is plainly more than 2.

The standard deviation has a check too. It is always about 29 percent of the width, and 29 percent of 2.5 is about 0.72. A value of 0.52 is the variance in disguise; a value of 1.15 would need a width of 4.

In context: a typical furnace repair takes 2.75 hours, give or take about three quarters of an hour.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 302 — Example 5.6(e)

64. Check: a conditional wait

Check

Now the reduced sample space. Decide which rectangle you are on before you compute.

Check your understanding

The bus wait is uniform between 0 and 15 minutes. Given that a person has already waited more than 5 minutes, what is the probability that the total wait is more than 10 minutes?

  • A. 0.5 (correct)
  • B. 0.333
  • C. 0.667
  • D. 0.222

Answer: A

Why: Knowing the wait has passed 5 minutes leaves a rectangle from 5 to 15, ten minutes wide, with height one tenth. The base from 10 to 15 is 5 minutes, so the probability is 5 times one tenth, which is one half.

Why B tempts people
That keeps the original height of one fifteenth: 5 times one fifteenth. It is the unconditional chance of waiting more than 10 minutes, and ignores what you were told.
Why C tempts people
That is the probability of waiting more than 5 minutes, 10 out of 15. It answers the given part of the question, not the question.
Why D tempts people
That multiplies the two probabilities, one third times two thirds. That would be the chance of both events if they were independent, and these are not.

By the second way: the probability of more than 10 is one third, the probability of more than 5 is two thirds, and one third divided by two thirds is one half. The two methods agree, as they always do.

In context: someone who has already waited 5 minutes has an even chance of waiting beyond the 10-minute mark. The wait so far does not make the bus due any sooner; within the remaining 10 minutes every moment is still equally likely.

The distractors each come from a real habit: forgetting to redraw, answering the condition instead of the event, and multiplying probabilities that are not independent.

65. Check: a percentile from the right

Check

Finally, a boundary. Sketch which side the area is on before choosing.

Check your understanding

A commute is uniform between 20 and 50 minutes. The slowest 20 percent of commutes take at least how long?

  • A. 44 minutes (correct)
  • B. 26 minutes
  • C. 40 minutes
  • D. 24 minutes

Answer: A

Why: The slowest fifth is the right-hand end. With width 30 minutes, the base from k to 50 must be one fifth of 30, which is 6 minutes, so k is 44 minutes. That boundary is the 80th percentile, since 80 percent of commutes are shorter.

Why B tempts people
That is the 20th percentile, 20 plus one fifth of 30. It shades the fastest fifth of commutes, on the left, instead of the slowest.
Why C tempts people
That is 0.8 times 50, measuring from 0 instead of from 20. It lands inside the interval by luck, but it is not the right boundary.
Why D tempts people
That is 0.8 times the width, 30, without adding the starting point of 20. It lies below the fastest possible commute.

The area equation is 50 minus k, times one thirtieth, equal to 0.2; multiplying by 30 gives 50 minus k equal to 6, so k is 44. Checking the left side: 44 minus 20 is 24, and 24 thirtieths is 0.8.

In context: one commute in five takes at least 44 minutes. The words slowest and at least both point to the right-hand tail, and the percentile you report counts the area on the left, so this is the 80th percentile.

Two of the distractors fail the bounds check before any arithmetic: a commute of 24 minutes is possible but cannot be where the slowest fifth begins, and a percentile computed from 0 ignores the 20 minutes every commute takes.

66. Explain the taller rectangle

Explain it to yourself

The conditional probability is the idea in this section that students find hardest to explain, even after they can compute it. Explaining it in your own words is the best test of whether it has landed.

Discussion prompt

In a few sentences, explain why, for the smiling times, the density given that the smile lasted more than 8 seconds has height one fifteenth instead of one twenty-third, and why that makes the probability of more than 12 seconds rise from about 0.48 to about 0.73.

A strong answer names three things: which values the condition removes, why the remaining ones are still equally likely, and why the total must stay one. If your explanation says only that conditioning makes probabilities bigger, go back to the counterexample slide, where it made one smaller.

A good extension is to say why the original a disappears from the answer. Once you know the smile passed 8 seconds, the fact that smiles could once have been as short as 0 is irrelevant; the new rectangle starts at 8.

67. Exit ticket: the oil change

Exit ticket

Finish Try It 5.6. The time to change the oil in a car is uniformly distributed between 11 and 21 minutes, X ~ U(11, 21), as you set up in Part 1.

Discussion prompt

Find the probability that an oil change takes more than 19 minutes, and the 50th percentile. Then find the mean and standard deviation, and say in a sentence what the 50th percentile means in context.

The first mark is the base: from 19 up to 21, two minutes, times one tenth. The second is the percentile equation, whose base starts at 11, not 0; writing k times one tenth equal to 0.5 gives 5 minutes, which is quicker than any oil change.

The last sentence earns the rest. A percentile is a time in minutes, and its meaning is a statement about what fraction of oil changes are faster. The fact that it equals the mean is a feature of flat distributions, and it will not survive into the next section.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, p. 302 — Try It 5.6, parts (d) and (e)

68. Recap

Recap

A uniform distribution describes a continuous variable that is equally likely to take any value in an interval from a to b. Its density is a rectangle whose height is forced by an area of one, so every question about it is arithmetic on a base and a height.

ideaformularead it as
notationX ~ U(a, b)equally likely from the lowest value a to the highest b
densityf(x) = 1/(b − a), a ≤ x ≤ bone over the width, so the area is one
probabilityP(c < x < d) = (d − c)/(b − a)the width you want as a share of the whole
one valueP(x = c) = 0endpoints never change a probability
meanμ = (a + b)/2the midpoint, also the median
standard deviationσ = (b − a)/√12about 29 percent of the width
percentilek = a + p(b − a)walk the proportion p along the interval
conditionalP(x > d | x > c) = (b − d)/(b − c)redraw on what is known, then take the share

The habits that carry the section: define X in words, draw the rectangle with a and b marked, keep every base inside it and starting at a, put a percentile's area on the side the words describe, and redraw the rectangle whenever you are told something. Then check: total area one, probabilities between 0 and 1, percentiles between a and b.

The Introductory Business Statistics text treats the same distribution in its §5.2, with business contexts, if you want more practice in a different setting. The next section, the exponential distribution, keeps the idea that probability is area but replaces the rectangle with a curve that falls away, where small values are more likely than large ones.

OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution §5.2, pp. 295-302 — the whole section

OpenStax Introductory Business Statistics 2e, §5.2 The Uniform Distribution §5.2 The Uniform Distribution — the same distribution in business contexts

Sources

  1. OpenStax Introductory Statistics 2e, §5.2 The Uniform Distribution — Illowsky & Dean, OpenStax / Rice University, CC BY 4.0, pp. 294-302
  2. OpenStax Introductory Business Statistics 2e, §5.2 The Uniform Distribution — Illowsky & Dean, OpenStax / Rice University, CC BY 4.0

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

Book on Wyzant · Text (657) 465-8108