6.1 The Standard Normal Distribution

The most important distribution in the book, and the foundation of everything from chapter 7 onward. A normal distribution is bell shaped and symmetric about its mean, and it is fixed by exactly two numbers: the mean, which slides the curve left or right, and the standard deviation, which makes it fatter or skinnier. Since those two may be anything, there are infinitely many normal distributions, and one of special interest is the standard normal, whose mean is zero and standard deviation one. The z-score converts any normal value into a standard one by measuring its distance from the mean in units of the standard deviation, which both reduces every normal distribution to a single one and makes values on completely different scales directly comparable. The Empirical Rule then reads off the areas that matter most: about 68, 95 and 99.7 percent of values lie within one, two and three standard deviations of the mean.

Subject: Statistics · 67 slides · symbolic lesson

Open the interactive version of this deck

What this lesson covers

The lesson, slide by slide

1. The Standard Normal Distribution

Title

Statistics · Chapter 6 — The Normal Distribution

Two numbers fix every bell curve, and the z-score turns all of them into one

2. What this lesson gives you

Objectives

Chapter 5 gave you continuous random variables, where a probability is an area under a curve. This chapter meets the most important curve of all, the bell-shaped normal curve, and this first section is about reading it: where it sits, how wide it is, and how to say where any value falls on it.

One idea carries the whole section: distance from the mean, measured in standard deviations. Once you can measure that, every normal distribution, whatever its units, becomes the same curve, and a height in centimetres can be compared with an exam score or a weight gain. Section 6.2 then uses that single curve to find probabilities.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, pp. 335-340 — the chapter introduction and the section this deck covers

3. Before anything new: a z-score for a data set

Warm-up

Section 2.7 gave you the z-score for a value in a data set: the distance from the mean divided by the standard deviation. Recall it now on a small case, because this section uses exactly the same idea for a whole distribution.

Discussion prompt

A class's quiz scores have mean 70 and standard deviation 8. You scored 82. How many standard deviations above the mean is your score? And what score would sit one standard deviation below the mean?

Keep both directions in mind: from a score to a number of standard deviations, and from a number of standard deviations back to a score. This section names the first one the z-score and writes the second as a formula, and almost every question in it is one or the other.

Also recall from Chapter 5 that for a continuous variable the total area under the curve is one, and a probability is an area. That fact is about to explain why a bell curve gets shorter when it gets wider.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 336 — a z-score is measured in units of the standard deviation

4. The two parameters

Section

Part 1

5. Where the bell curve comes from

Concept

The book's collaborative exercise asks your class to record heights and draw a histogram. Here is what that looks like with a big class: 200 heights, simulated from the Chilean data you will meet in Example 6.3, with a smooth curve drawn through the bars.

Figure (svg): A histogram of 200 simulated heights in bins 2.5 cm wide from 150 to 190 cm, highest in the middle near 170 and falling away on both sides. A smooth yellow bell curve is drawn over the bars. The sample mean is 170.0 cm and the sample standard deviation 6.44 cm.

The bars are data; the yellow curve is the model. It is fixed by two numbers only, the centre and the spread, and it is what the bars would settle into with thousands of heights instead of 200.

The histogram is lumpy, because 200 is not very many, but its shape is unmistakable: one peak in the middle, falling away symmetrically on both sides. The sample mean came out at 170.0 cm and the sample standard deviation at 6.44 cm, close to the 170 and 6.28 the heights were drawn from.

normal distribution — A continuous probability distribution whose graph is a symmetric, bell-shaped curve. It is described completely by two parameters, its mean μ and its standard deviation σ.

The book is honest that the normal is widely used and even more widely abused: shoe sizes, heights, IQ scores and measurement errors are close to normal, but plenty of real data are skewed and are not. When you use the normal you are assuming the data are shaped like this picture.

Why does height come out bell-shaped? A height is the sum of many small, roughly independent influences, from many genes and many years of diet, and sums of many small influences tend to pile up in the middle and thin out symmetrically. Chapter 7 makes that idea precise. For now, the picture is the working definition: one central peak, symmetric shoulders, thin tails on both sides.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, pp. 335-336 — chapter introduction and collaborative exercise

6. Two numbers are all a normal curve needs

Concept

The normal distribution has two parameters, two numerical descriptive measures: the mean μ and the standard deviation σ. When a quantity X has a normal distribution with those two, you write it like this.

\[ X \sim N(\mu, \sigma) \]

Read the tilde as is distributed as. In the Chilean heights, μ is 170 cm and σ is 6.28 cm, and the picture below shows where each number lives on the curve.

Figure (svg): A bell curve over heights from 148 to 192 cm, peaked at 170. A dashed yellow line rises from 170 to the peak. An orange arrow runs horizontally from 170 to 176.28, labelled sigma equals 6.28, ending at an orange dot where the curve stops bending downward; a matching dot sits at 163.72.

The mean is where the curve peaks and folds in half. The standard deviation is the distance from the centre to either point where the curve changes from bending down to bending up.

The mean is the centre: the curve is symmetric about a vertical line through μ, so the mean and the median are the same number. The standard deviation is the width: it is the distance from the centre to the point where the curve stops bending downward and starts bending up. That is a useful way to read σ straight off a picture.

The density formula behind the curve is complicated, and the book tells you plainly not to memorise it. You never need it here: every calculation in this section uses only μ and σ.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, pp. 335-336 — X ~ N(μ, σ) and the symmetry about μ

7. Decoding X ~ N(μ, σ)

Notation

Short notation, five pieces. Each one is worth reading slowly, because the order of the two numbers catches people out when they move between books.

Annotate

On: \( X \sim N(\mu,\ \sigma) \)

  • The random variable: the quantity being measured, such as the height of one randomly chosen boy. Always define it in words first.
  • Read as is distributed as, or follows.
  • The family: normal. Other letters name other families, such as U for uniform and P for Poisson.
  • The first number is the mean, the centre of the curve.
  • The second number is the standard deviation, in the same units as X. In this book it is never the variance.

That last note matters. Some texts put the variance, σ squared, in the second place. In this book, and in this deck, the second number is always the standard deviation, so X ~ N(1, 2) has a standard deviation of 2, not of the square root of 2.

\[ X \sim N(5, 6): \quad \mu = 5, \quad \sigma = 6 \]

The standard deviation can be larger than the mean, as it is here; nothing forbids it. It simply means a good share of the curve lies below zero.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 335 — the notation X ~ N(μ, σ)

8. Changing μ slides the curve

Concept

Hold σ fixed at 6 and change only the mean. The book says a change in μ causes the graph to shift to the left or right, and that is all it does.

Figure (svg): Three identical bell curves, each with standard deviation 6, centred at 35, 50 and 65. Their peaks have the same height and their widths are the same; yellow arrows point left and right from the middle curve.

Changing μ slides the whole curve along the axis. Nothing about its shape changes: same width, same height.

\[ N(35, 6) \quad N(50, 6) \quad N(65, 6) \]

The three curves are the same shape and the same height; each is a copy of the others moved along the axis. The mean is a location parameter: it says where the curve sits, not what it looks like.

So a class whose average height is 5 cm more than another's, with the same spread, has exactly the same curve, pushed 5 cm to the right. Every percentage about the class, the share within one standard deviation, the share above the mean, is unchanged by the move.

A useful way to hold this: adding the same number to every value in a data set moves the mean by that number and leaves the standard deviation alone, as you saw in Chapter 2. A normal distribution behaves the same way, because it is the shape the data would take with endlessly many values.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 336 — a change in μ shifts the graph

9. Changing σ reshapes it, and the area stays one

Concept

Now hold the mean at 50 and change only σ. The book says the curve becomes fatter or skinnier depending on σ, and it gives the reason: the area under the curve must equal one.

Figure (svg): Three bell curves all centred at 50. The one with sigma 4 is tall and narrow with peak height 0.100, sigma 8 is medium with peak 0.050, and sigma 16 is low and wide with peak 0.025. Dotted lines lead from each peak to its label.

Doubling σ doubles the width and halves the height. It has to: each curve encloses an area of exactly one.

Follow the peaks. The peak height of a normal curve is one over σ times the square root of two pi, so it is inversely proportional to σ.

\[ \text{peak} = \frac{1}{\sigma\sqrt{2\pi}} \]

\[ \sigma = 4: 0.100, \quad \sigma = 8: 0.050, \quad \sigma = 16: 0.025 \]

Double the width and the height must halve, or the area would double. You do not need the formula to see this; you need only the fact from Chapter 5 that a density's total area is one. A wider curve is a lower curve.

Since μ can be any number and σ any positive number, there are infinitely many normal distributions. That sounds like a problem for anyone who has to compute areas, and Part 3 shows why it is not.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 336 — a change in σ changes the shape; infinitely many normal distributions

10. Move the two parameters yourself

Tweak it

The slider draws the curve X ~ N(μ, σ) for any mean and standard deviation you choose. Try one parameter at a time, and watch what changes and what does not.

Parameter explorer

Start at mean 0 and standard deviation 1 and read the peak height. Then set the standard deviation to 2 and read it again. Finally move the mean to 5 without touching the standard deviation, and say what changed.

\[ X \sim N({m},\ {s}) \]

  • m — from -6 to 6: mean μ
  • s — from 0.5 to 4: standard deviation σ

\[ \sigma = 1: \text{ peak} \approx 0.399, \qquad \sigma = 2: \text{ peak} \approx 0.199 \]

Notice that the peak never moves off the mean, whatever you do to σ, and that the curve never tips to one side: every normal curve is symmetric. At σ equal to one half the curve shoots up to about 0.8, taller than one. That is allowed: a density is not a probability, and only the area has to equal one.

11. Which parameter changed?

Sorting

Each distribution below is compared with N(50, 10). Decide whether the curve slid, changed width, or both. Remember that the second number is the standard deviation.

Sort into buckets

Sort each distribution by how its curve differs from N(50, 10).

slid only (same shape)
N(70, 10); N(20, 10)
same centre, new width
N(50, 3); N(50, 25)
slid and changed width
N(60, 5); N(35, 20)
slide
The standard deviation is still 10, so the shape is identical; only the centre moved, right to 70 or left to 20.
shape
The mean is still 50, so the peak stays put; σ of 3 makes a tall narrow curve and σ of 25 a low wide one.
both
Both numbers differ from 50 and 10, so the curve moves and changes width at once.

Reading the pair in order is the whole skill: first number, where; second number, how wide. N(50, 3) and N(50, 25) sit in exactly the same place and look nothing alike, while N(70, 10) and N(20, 10) are far apart but identical in shape.

If you put N(50, 25) with the slides, you probably read the 25 as a new centre. It is not; the centre is always the first number.

12. What happens to the peak when σ triples?

Prediction

A machine's fill volumes follow N(50, 5). A worn part makes them follow N(50, 15) instead. Before you look at any picture, predict what the new curve looks like.

Predict first

Compared with N(50, 5), how does the curve of N(50, 15) look?

  • three times as wide and one third as tall, still centred at 50
  • the same shape, moved 10 to the right
  • three times as wide and just as tall
  • narrower and taller, since the spread number went up

Correct: three times as wide and one third as tall, still centred at 50

Why: The mean is still 50, so the curve does not move. The standard deviation triples, so the curve is three times as wide, and because the area must stay equal to one, the peak drops to one third of its height.

\[ \frac{1}{15\sqrt{2\pi}} = \frac{1}{3} \cdot \frac{1}{5\sqrt{2\pi}} \]

The option with the same height is the tempting one, and it is impossible: a curve three times as wide at the same height would enclose about three times the area. For the machine, a worn part means far more bottles far from 50, even though the average fill has not changed at all.

13. Trap: reading σ as a variance

Trap

The trap

Practice Exercise 3 gives X ~ N(1, 2) and asks for σ. A first attempt, remembering a formula from another book, writes:

\[ \sigma^{2} = 2 \]

\[ \sigma = \sqrt{2} \approx 1.41 \]

Wrong for this book. The second number is already σ.

The fix

In this book the notation always puts the standard deviation second. Nothing needs to be square-rooted.

\[ X \sim N(1, 2) \]

\[ \sigma = 2 \]

Some texts and some software write the variance in the second place instead, so check the convention whenever you read a normal distribution from somewhere else. Within this course, N(μ, σ) is the rule, and the book's answer to Exercise 3 is 2.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 practice, p. 351 — Exercise 3

14. Symmetry makes the mean the median

Concept

Because the curve is symmetric about μ, exactly half of the area lies on each side of the mean. The value with half the distribution below it is, by definition, the median.

\[ P(X < \mu) = P(X > \mu) = 0.5 \]

\[ \text{median} = \mu \]

That answers two of the book's practice exercises at a glance. A normal distribution with mean 61 and standard deviation 15 has median 61, and X ~ N(minus 4, 1) has median minus 4. The standard deviation plays no part.

\[ N(61, 15): \text{ median} = 61 \qquad N(-4, 1): \text{ median} = -4 \]

The same symmetry is behind the homework question about surgical recovery times with mean 5.3 days: the median is 5.3 days too. For skewed data the mean and median separate, and seeing them far apart in real data is one sign that a normal model does not fit.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 practice and homework, pp. 351, 355 — Exercises 2, 5 and 60

15. The z-score

Section

Part 2

16. Distance from the mean, counted in standard deviations

Concept

The book opens the section with a small example. The mean is 5 and the standard deviation is 2. Where does 11 sit? Count steps of σ from the mean.

Figure (svg): A bell curve for mean 5 and standard deviation 2, with x ticks at minus 1, 1, 3, 5, 7, 9 and 11 and a second row of z values minus 3 to 3 underneath. Three orange arrows, each labelled plus sigma, step from 5 to 7, 7 to 9 and 9 to 11, where a yellow line marks x equals 11.

Three steps of σ from the mean lands on 11, so the z-score of 11 is 3. The orange row is the same axis counted in standard deviations.

\[ 11 = 5 + (3)(2) \]

Three steps of 2 from 5 lands on 11, so 11 is three standard deviations above the mean, and its z-score is 3. A z-score is simply that count: how many standard deviations a value sits from the mean.

\[ x = \mu + (z)(\sigma) \]

Look at the orange row under the axis in the picture. It is the same axis relabelled in standard deviations. The value 5 becomes 0, the value 11 becomes 3, and the value 1 becomes minus 2. A negative count simply means you walked left.

Think of σ as a ruler whose length is set by the distribution itself. With a standard deviation of 2, each step is 2 units; with a standard deviation of 6, each step is 6. The z-score is the number of ruler-lengths, and that is why it means the same thing on every scale.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 336 — the opening example: 5 + (3)(2) = 11

17. Solving for z

Concept

The book defines the z-score by a formula. You can get it in two moves from the pattern on the last slide, which is worth doing once so the formula never feels arbitrary.

\[ x = \mu + z\sigma \]

First move: subtract μ from both sides. The left side is now the distance of x from the mean, in the original units, with a sign.

\[ x - \mu = z\sigma \]

Second move: divide both sides by σ. Dividing a distance by the size of one standard deviation counts how many standard deviations fit into it.

\[ z = \frac{x - \mu}{\sigma} \]

So the z-score and the formula x equals μ plus zσ are one relationship solved two ways. Use the first when you know the value and want its position; use the second when you know the position and want the value. Parts 2 and 4 use both constantly.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 336 — the z-score definition

18. Decoding z = (x − μ)/σ

Notation

Every piece of the formula has a job. Read them one at a time.

Annotate

On: \( z = \frac{x - \mu}{\sigma} \)

  • One particular value of the variable, such as a height of 168 cm.
  • How far x is from the mean, in the original units. Its sign is the direction: positive to the right of the mean, negative to the left.
  • Converts that distance into standard deviations. The units cancel, so z has no units at all.
  • The answer: how many standard deviations x is above or below the mean. z = 0 means x equals the mean.

Because the units cancel, a z-score of 2 means the same thing for heights in centimetres, weights in pounds and scores in points. That is the property Part 4 exploits.

\[ x > \mu \Rightarrow z > 0, \quad x < \mu \Rightarrow z < 0, \quad x = \mu \Rightarrow z = 0 \]

Use those three facts as a check every time. Before you compute, ask whether x is above or below the mean; the sign of your answer must agree.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 336 — values above the mean have positive z-scores

19. Example 6.1: two z-scores in N(5, 6)

Worked example

Suppose X ~ N(5, 6), so the mean is 5 and the standard deviation is 6. Find the z-scores of x equal to 17 and of x equal to 1, and say what each one means.

\[ X \sim N(5, 6), \quad x = 17 \text{ and } x = 1 \]

Substitute x = 17

Why: above the mean, so expect a positive z.

\[ z = \frac{17 - 5}{6} \]

Subtract

Why: 17 is 12 units above the mean.

\[ z = \frac{12}{6} \]

Divide

Why: 12 units is two standard deviations of 6.

\[ z = 2 \]

So 17 is two standard deviations above, or to the right of, the mean 5. Now the second value, which lies below the mean, so this time the sign should come out negative.

Substitute x = 1

Why: below the mean, so expect a negative z.

\[ z = \frac{1 - 5}{6} \]

Subtract

Why: 1 is 4 units below the mean.

\[ z = \frac{-4}{6} \]

Divide and round to two places

Why: four sixths is two thirds.

\[ z \approx -0.67 \]

Figure (svg): A bell curve for mean 5 and standard deviation 6 with x ticks from minus 13 to 23 in steps of 6 and z values minus 3 to 3 underneath. A yellow line at x equals 17 sits exactly on the z equals 2 tick. A pink line at x equals 1 sits two thirds of the way from 5 back to minus 1, marked minus 0.67 on the z row.

17 lands on a tick of the z row, 2. The value 1 falls between ticks, a little over half a standard deviation to the left of 5, so its z-score is a fraction.

So 1 is 0.67 standard deviations below, or to the left of, the mean 5. On the picture, 17 lands exactly on a tick of the z row while 1 falls between the ticks for 0 and minus 1, closer to minus 1: a fractional z-score is perfectly normal, and most values have one.

The book summarises the pattern in one line: when z is positive, x is above the mean, and when z is negative, x is below it. It then checks each answer by rebuilding x from μ plus zσ, and so should you.

Check by rebuilding each value

Why: 17 comes back exactly. The book says the second rebuild is approximately one; it is 0.98, and the gap is only the rounding of two thirds to 0.67, not an error.

\[ 5 + (2)(6) = 17, \qquad 5 + (-0.67)(6) = 0.98 \approx 1 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, pp. 336-337 — Example 6.1

20. Try It 6.1: a value far below the mean

Fill the middle

What is the z-score of x equal to 1 when X ~ N(12, 3)? Fill in the numerator and the result, to two decimal places. Decide the sign before you compute.

Fill in the blanks

z = \frac-11-3.67 = \frac___}___ \approx ___

Why: One is 11 below the mean of 12, so the numerator is minus 11. Eleven thirds is about 3.67, and the value is below the mean, so the z-score is minus 3.67.

Minus 3.67 means the value 1 is almost three and two thirds standard deviations below the mean. Part 5 will tell you how rare that is: less than 0.15 percent of a normal distribution lies that far out on the left.

If you wrote positive 3.67, you subtracted the wrong way round, 12 minus 1. The order in the formula is always the value minus the mean, so that values below the mean come out negative.

\[ 12 + (-3.67)(3) = 0.99 \approx 1 \]

The rebuild lands on 0.99, one hundredth away from 1 only because of rounding: the check passes.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 337 — Try It 6.1

21. Example 6.2(a)(b): losing and gaining weight

Worked example

Some doctors believe a person can lose five pounds a month, on average, by cutting fat and exercising. Suppose weight loss is normal with a standard deviation of two pounds, so X ~ N(5, 2), where X is the pounds lost in a month.

\[ X \sim N(5, 2) \]

(a) Substitute a loss of 10 pounds

Why: above the mean of 5.

\[ z = \frac{10 - 5}{2} \]

Subtract, then divide

Why: 5 pounds is two and a half standard deviations.

\[ z = \frac{5}{2} = 2.5 \]

So 10 pounds is 2.5 standard deviations to the right of the mean, five pounds. Now part (b): a person gained three pounds. A gain is a negative loss, so x is minus 3, and it goes into the formula with its sign.

(b) Substitute a loss of minus 3

Why: a gain of 3 pounds is a loss of minus 3 pounds.

\[ z = \frac{-3 - 5}{2} \]

Subtract

Why: minus 3 is 8 pounds below the mean.

\[ z = \frac{-8}{2} \]

Divide

Why: 8 pounds is four standard deviations.

\[ z = -4 \]

Figure (svg): A bell curve for weight lost with mean 5 and standard deviation 2, x from minus 4.5 to 14.5, z row minus 4 to 4 underneath. A yellow line at 10 pounds, z equals 2.5, sits in the right tail. A red line at minus 3, z equals minus 4, sits where the curve has almost reached the axis.

A weight gain is a negative weight loss, so it sits to the left of zero on this axis. Four standard deviations below the mean is so far out that the curve is almost flat on the axis there.

So a gain of three pounds is four standard deviations to the left of the mean. On the picture it sits where the curve has all but reached the axis: under this model, gaining weight while following the plan should almost never happen.

That is a good moment to question the model rather than the person. If someone following the plan really did gain three pounds, either something unusual happened or weight change is not normal with this mean and spread. A z-score of minus 4 is evidence worth taking seriously, and later chapters turn exactly this kind of reasoning into hypothesis tests.

Check by rebuilding both values

Why: μ plus zσ must give back 10 and minus 3.

\[ 5 + (2.5)(2) = 10, \qquad 5 + (-4)(2) = -3 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 337 — Example 6.2(a) and (b)

22. Four attempts at a z-score for 168 cm

Error analysis

Example 6.3 uses X ~ N(170, 6.28) for the heights of 15 to 18-year-old males in Chile in 2009 to 2010. Four students found the z-score of a boy 168 cm tall. Only one is right. Find the fault in each of the others.

Annotate

On: \( \frac{170 - 168}{6.28} \quad \frac{168 - 170}{170} \quad 168 - 170 \quad \frac{168 - 170}{6.28} \)

  • Subtracts in the wrong order. The size is right but the sign says the boy is taller than average, and he is shorter.
  • Divides by the mean instead of the standard deviation. That is a proportional difference, about 1 percent, not a count of standard deviations.
  • Stops halfway. Minus 2 is the distance in centimetres; it still has to be divided by σ to become a z-score.
  • Correct: value minus mean, over the standard deviation. The boy is 0.32 standard deviations below the mean.

The first error is the dangerous one, because its size matches the book's answer. Only the direction check catches it: 168 is less than 170, so the z-score must be negative. Say above or below before you compute, every time.

The second and third errors both come from forgetting what the denominator is for. Dividing by σ is what turns centimetres into standard deviations; without it, or with the wrong divisor, the number has no meaning on the z row.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 338 — Example 6.3(a)

23. Example 6.3(a): a boy 168 cm tall

Worked example

Let X be the height of a 15 to 18-year-old male from Chile in 2009 to 2010. Male heights are known to be normal, with mean 170 cm and standard deviation 6.28 cm. Find and interpret the z-score of 168 cm.

\[ X \sim N(170, 6.28), \quad x = 168 \]

Predict the sign

Why: 168 is below 170, so z must be negative.

\[ 168 < 170 \Rightarrow z < 0 \]

Substitute

Why: value minus mean, over σ.

\[ z = \frac{168 - 170}{6.28} \]

Subtract

Why: 2 cm below the mean.

\[ z = \frac{-2}{6.28} \]

Divide and round

Why: node gives minus 0.3185.

\[ z \approx -0.32 \]

Figure (svg): The height curve for mean 170 and standard deviation 6.28, with ticks one standard deviation apart from 151.16 to 188.84 and z values minus 3 to 3 beneath. A pink line at 168 cm stands just left of the peak; a short pink arrow from 170 to 168 is labelled 2 centimetres equals 0.32 sigma.

168 is only 2 cm below 170, and a standard deviation here is 6.28 cm, so it is about a third of a standard deviation to the left: an entirely ordinary height.

Interpret it in context: a height of 168 cm is 0.32 standard deviations to the left of the mean height of 170 cm. That fills in all four of the book's blanks: minus 0.32, 0.32, left, 170. A third of a standard deviation is a very ordinary distance from the mean.

Notice how the interpretation sentence is built: the size of z becomes the number of standard deviations, the sign becomes the side, and the mean is named with its units. Write that sentence every time; a bare minus 0.32 answers only half the question.

Check by rebuilding x

Why: μ plus zσ with the rounded z comes back to 168 within rounding.

\[ 170 + (-0.32)(6.28) = 167.99 \approx 168 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 338 — Example 6.3(a)

24. Example 6.3(b): from a z-score back to a height

Worked example

Now the other direction. A boy from the same population has a z-score of 1.27. What is his height? Here z is known and x is not, so use the formula that is solved for x.

\[ x = \mu + z\sigma \]

Substitute

Why: mean 170, z of 1.27, σ of 6.28.

\[ x = 170 + (1.27)(6.28) \]

Multiply

Why: 1.27 standard deviations, converted to centimetres.

\[ x = 170 + 7.9756 \]

Add and round to two places

Why: the book's answer.

\[ x = 177.9756 \approx 177.98 \text{ cm} \]

So the boy is about 177.98 cm tall, and the z-score says he is 1.27 standard deviations to the right of the mean. The product zσ is the step people skip: it is the distance from the mean in centimetres, and it is what gets added.

Figure (svg): The height curve for mean 170 and standard deviation 6.28. A yellow band from 170 to 177.98 ends at a line labelled z equals 1.27; inside it the distance plus 7.98 is written. An orange band from 157.44 to 170 starts at a line labelled z equals minus 2; inside it the distance minus 12.56 is written.

Going backwards, z times σ is the distance in centimetres, and its sign says which way to walk from 170.

The picture also shows Try It 6.3(b), a z-score of minus 2: the product is minus 12.56 cm, so that boy is 157.44 cm tall. The sign of z decides which way you walk from 170.

Check by computing z from the answer

Why: the height should give back 1.27.

\[ \frac{177.98 - 170}{6.28} = \frac{7.98}{6.28} \approx 1.27 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, pp. 338-339 — Example 6.3(b) and Try It 6.3(b)

25. Try It 6.3: which formula, before any arithmetic?

Step zero

Use the Chilean heights, X ~ N(170, 6.28). Part (a) gives a height of 176 cm and asks for its z-score. Part (b) gives a z-score of minus 2 and asks for the height. Before computing either, decide which formula each part needs, and what sign or side you expect.

Discussion prompt

For each part, name the formula you would use and predict the side of the mean the answer is on. Then carry both out and interpret each answer in a sentence.

The first move in every problem of this section is exactly this choice. Look at what the question hands you. A value means you want z; a number of standard deviations means you want x.

Predicting the side first is not wasted effort. It takes a second, and it catches both of the common sign errors: subtracting the wrong way round in part (a), and adding 12.56 instead of subtracting it in part (b).

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 339 — Try It 6.3

26. Try It 6.2: Jerome's ten-point game

Prediction

Jerome averages 16 points a game with a standard deviation of four points, so X ~ N(16, 4). In one game he scores ten points, and the z-score is minus 1.5. Choose the sentence that fills the book's blanks correctly.

Predict first

What does a z-score of minus 1.5 say about Jerome's ten-point game?

  • 10 is 1.5 standard deviations to the left of the mean 16
  • 10 is 1.5 points below the mean 16
  • 10 is 1.5 standard deviations to the right of the mean 16
  • 10 is 1.5 times smaller than the mean 16

Correct: 10 is 1.5 standard deviations to the left of the mean 16

Why: The z-score counts standard deviations, not points, and its minus sign says below, or to the left of, the mean. Ten is six points below 16, and six points is one and a half standard deviations of four.

\[ z = \frac{10 - 16}{4} = \frac{-6}{4} = -1.5 \]

Each wrong sentence misreads one part of the number. Reading 1.5 as points forgets the division by σ: he actually scored six points under his average. Reading it as to the right ignores the sign. And a z-score is never a ratio of the value to the mean.

The last option is worth a second look because it is the mistake in the book's own homework Exercise 65, where one of the offered readings of a z-score of 1.75 is 1.75 times the average. A z-score compares a distance with σ, never a value with μ.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 338 — Try It 6.2

27. Read each z-score as a sentence

Matching

Practice Exercises 26 to 30 give a value and its z-score and ask for the number of standard deviations and the side. Match each pair with its sentence.

Match the pairs

  • a. x = 5, z = −1.25
  • b. x = 3, z = 0.67
  • c. x = −2, z = 6
  • d. x = −5, z = −3.14
  • e. x = 6, z = −1.7
  • v. 1.25 standard deviations to the left of the mean
  • w. 0.67 standard deviations to the right of the mean
  • x. 6 standard deviations to the right of the mean
  • y. 3.14 standard deviations to the left of the mean
  • z. 1.7 standard deviations to the left of the mean

Why: Only the z-score matters. Its size is the number of standard deviations and its sign is the side: negative means left of the mean, positive means right. The value of x itself says nothing about the side.

The pairs that catch people are the ones where the sign of x and the sign of z disagree. A value of minus 2 with z equal to 6 is six standard deviations to the right: the mean must be well below minus 2. A value of 6 with z equal to minus 1.7 is to the left: the mean is above 6.

So never read the side off the value. The side is relative to the mean, and only z carries that information.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 practice, p. 352 — Exercises 26-30

28. Trap: dropping the sign of z

Trap

The trap

Try It 6.3(b): a boy has z equal to minus 2. A first attempt treats two standard deviations as a distance to add:

\[ x = 170 + (2)(6.28) \]

\[ x = 182.56 \text{ cm} \]

Wrong. The boy is taller than average, but z said below.

The fix

The sign of z is the direction. A negative z-score must give a value below the mean.

\[ x = 170 + (-2)(6.28) \]

\[ x = 170 - 12.56 \]

\[ x = 157.44 \text{ cm} \]

Keep z with its sign inside the formula and the arithmetic handles the direction for you. Then check: 157.44 is below 170, as a negative z-score demands.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 339 — Try It 6.3(b)

29. Homework 63(c): would you believe a z-score of 3.5?

Estimation

Heights of NBA players are roughly normal with mean 79 inches and standard deviation 3.89 inches. A player says his height has a z-score of 3.5. Before computing, estimate how unusual that is.

Predict first

How believable is a z-score of 3.5 for an NBA player's height?

  • very doubtful: it is beyond three standard deviations, where almost nobody is
  • ordinary: about one player in six is that tall
  • impossible: no z-score can be above 3
  • believable: a z-score of 3.5 is only 3.5 inches above average

Correct: very doubtful: it is beyond three standard deviations, where almost nobody is

Why: Only about 0.15 percent of a normal distribution lies more than three standard deviations above the mean, and 3.5 is further out still. It is not impossible, but it would make him taller than virtually every player in the league.

\[ x = 79 + (3.5)(3.89) = 92.615 \text{ inches} \]

That is about seven feet eight and a half inches, taller than the tallest players the NBA has ever listed. A z-score of 3.5 is rare but not impossible under the model; with a real, well-documented height distribution, you would want very good evidence before believing it.

The option that says 3.5 inches is the Jerome mistake again: z counts standard deviations, and here each one is 3.89 inches, so the claim is 13.6 inches above the average.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 homework, pp. 355-356 — Exercise 63(c)

30. Homework 63(a)(b): two NBA heights

Worked example

NBA heights: X ~ N(79, 3.89) in inches, from 430 players on team rosters. Calculate the z-scores of 77 inches and 85 inches and interpret each in a complete sentence, as the exercise asks.

\[ X \sim N(79, 3.89) \]

(a) Substitute x = 77

Why: below the mean, so expect a negative z.

\[ z = \frac{77 - 79}{3.89} \]

Subtract and divide

Why: 2 inches is about half a standard deviation.

\[ z = \frac{-2}{3.89} \approx -0.51 \]

So a player 77 inches tall is about 0.51 standard deviations below the mean height of 79 inches.

(b) Substitute x = 85

Why: above the mean.

\[ z = \frac{85 - 79}{3.89} \]

Subtract and divide

Why: 6 inches is about one and a half standard deviations.

\[ z = \frac{6}{3.89} \approx 1.54 \]

So a player 85 inches tall is about 1.54 standard deviations above the mean height of 79 inches.

Figure (svg): The curve of NBA heights with mean 79 inches and standard deviation 3.89, ticks one standard deviation apart from 67.3 to 90.7 and z values minus 3 to 3 beneath. Pink line at 77 inches just left of centre, yellow line at 85 inches about one and a half standard deviations right, and a red line far out at 92.6 inches where the curve has nearly reached the axis.

77 and 85 are ordinary heights for a basketball player. 92.6 inches, over seven foot eight, is further out than the whole z row of the picture.

Both are unremarkable heights for this population, well within two standard deviations. The claimed z-score of 3.5 is in another league entirely, which is why the picture needed a wider window to show it.

Check by rebuilding both heights

Why: the rounded z-scores give back 77 and 85 to within rounding.

\[ 79 + (-0.51)(3.89) = 77.02, \quad 79 + (1.54)(3.89) = 84.99 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 homework, pp. 355-356 — Exercise 63(a)(b)

31. The standard normal

Section

Part 3

32. Standardising: subtract μ, then divide by σ

Concept

Now apply the z-score formula not to one value but to every value of X at once. The book's first sentence of the section says what comes out: the standard normal distribution is a normal distribution of standardised values called z-scores.

Figure (svg): Three panels on the same horizontal scale. Left: a low wide curve centred at 5, labelled X is N of 5 and 2. Middle: the same curve moved so it is centred at 0, labelled X minus 5 is N of 0 and 2. Right: a curve centred at 0 that is half as wide and twice as tall, labelled X minus 5 over 2 is N of 0 and 1. Arrows between the panels read subtract 5 and divide by 2.

Subtracting the mean only slides the curve. Dividing by σ squeezes it to width one, and to keep the area equal to one it has to grow taller.

Watch the two moves in the picture. Subtracting the mean slides every value left by 5, so the whole curve is now centred at zero; its width has not changed. Dividing by σ then shrinks every distance from zero by a factor of 2, so the curve becomes half as wide, and twice as tall to keep its area.

\[ X \sim N(5, 2) \quad \longrightarrow \quad Z = \frac{X - 5}{2} \sim N(0, 1) \]

Nothing in this argument depended on 5 and 2. Any normal distribution, standardised this way, becomes the same curve with mean zero and standard deviation one.

The order of the moves matters. Subtract first, then divide: that is what the brackets in the formula say. Divide first and you get X over 2 minus 5, which moves the curve to a different place entirely, centred at minus 2.5 rather than zero.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 336 — the transformation z = (x − μ)/σ produces Z ~ N(0, 1)

33. Why the mean is 0 and the standard deviation is 1

Concept

You can see both numbers without any calculus, from what the two moves do. Take the mean first. If every value has μ subtracted, the average of the new values is the old average minus μ.

\[ \text{mean of } (X - \mu) = \mu - \mu = 0 \]

Dividing by σ keeps a zero mean at zero. Now the spread. Subtracting a constant slides values without spreading them, so X minus μ still has standard deviation σ. Dividing every value by σ divides every distance by σ, including the standard deviation itself.

\[ \text{sd of } (X - \mu) = \sigma \]

\[ \text{sd of } \frac{X - \mu}{\sigma} = \frac{\sigma}{\sigma} = 1 \]

standard normal distribution — The normal distribution with mean 0 and standard deviation 1, written Z ~ N(0, 1). It is the distribution of the z-scores of any normal variable.

The book's Exercise 9 asks what standardising does to the mean, and the answer is on the first line: it becomes zero. Its partner, what it does to the standard deviation, is on the last line: it becomes one.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 336; practice p. 352 — Z ~ N(0, 1); Exercise 9

34. Every normal curve is one curve with a different ruler

Picture it

Here is the payoff of standardising, as a picture. Draw the standard normal curve once. Under its tick marks, write the values that sit at each z-score for three different normal distributions from this section.

Figure (svg): One standard normal curve drawn once, with four rows of labels under its seven tick marks. The first row reads minus 3 to 3; the second, for X distributed N of 5 and 6, reads minus 13 to 23; the third, for Y distributed N of 2 and 1, reads minus 1 to 5; the fourth, for heights N of 170 and 6.28, reads 151.2 to 188.8. A dot on the curve at z equals 2 is labelled 17 in X, 4 in Y, 182.56 centimetres.

Every normal distribution is this one curve with a different ruler under it. The z row is the ruler they all share.

Read the picture column by column. Each column is one z-score, and underneath it are the values that sit that many standard deviations from their own mean in each distribution.

Every row lines up. In X ~ N(5, 6), the value 17 sits under z equal to 2; in Y ~ N(2, 1), the value 4 does; for the Chilean heights, 182.56 cm does. The curve above them is the same curve, so the share of each distribution beyond that point is the same too.

That is why the book says there are infinitely many normal distributions and then is untroubled by it. You never need a separate table or picture for each one. You translate into the z row, work with the one standard curve, and translate back.

\[ \text{value} \xrightarrow{\ z = (x-\mu)/\sigma \ } \text{standard curve} \xrightarrow{\ x = \mu + z\sigma \ } \text{value} \]

Section 6.2 does exactly this to find probabilities, and Chapters 7 to 10 do it for every test and interval. Everything rests on this one picture.

35. Is every mean-zero normal standard?

Counterexample

Practice Exercise 10 asks whether X ~ N(0, 1) is a standardised normal distribution. It is worth pushing further: which conditions are really required, and is one of them enough?

Discussion prompt

Is X ~ N(0, 1) a standard normal distribution, and why? Then give a normal distribution with mean zero that is not standard, and one with standard deviation one that is not standard, and say what goes wrong in each.

Both conditions are needed. A distribution is standard only when its values and its z-scores are the same numbers, and that requires the centre to be zero and the unit of spread to be one at the same time.

A quick test: pick any value and compute its z-score. For N(0, 1) you get the value back unchanged. For N(0, 2) or N(3, 1) you do not, so those distributions are not yet standardised.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 practice, p. 352 — Exercise 10

36. Growth charts: z-scores in a clinic

Real world

The World Health Organization tracks child development by comparing a child's weight with other children of the same height and sex. In 2009, weights of 80 cm girls in the reference population had X ~ N(10.2, 0.8) in kilograms. Doctors report the result as a z-score.

Discussion prompt

Homework Exercise 66: calculate and interpret the z-scores of 11 kg, 7.9 kg and 12.2 kg for an 80 cm girl. Which girl, if any, would a doctor want to look at more closely, and why is a z-score more useful to the doctor than the raw weight?

This is a real use: the WHO's cut-offs for underweight and overweight in children are written as z-scores, typically minus 2 and plus 2. That is the Empirical Rule of Part 5 being used as a screening test.

Notice that the distance matters more than the kilograms. The lighter girl is only 2.3 kg below the mean, but with a standard deviation of 0.8 kg that is far out in the tail.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 homework, p. 356 — Exercise 66

37. Comparing across scales

Section

Part 4

38. Raw values from different curves cannot be compared

Picture it

Two groups of people are trying to gain weight over six weeks. In the first group, gains are X ~ N(5, 6) in pounds; in the second, Y ~ N(2, 1). One person from the first group gains 17 pounds; one from the second gains 4. Who did better, relative to their own group?

Figure (svg): Two curves on one axis of pounds gained. X, distributed N of 5 and 6, is low and wide; Y, distributed N of 2 and 1, is tall and narrow around 2. The tail of X beyond 17 is shaded blue and the tail of Y beyond 4 is shaded orange; a note says each shaded tail is 2.28 percent of its group.

On the raw axis 17 and 4 look nothing alike. Measured against their own curves they are the same: two standard deviations up, with the same small share of the group beyond them.

On the raw axis, 17 looks far more impressive than 4. But the first group's gains are spread widely, and 17 is 12 pounds above their mean; the second group's gains are packed tightly, and 4 is 2 pounds above theirs. The fair comparison is how far each is from its own mean in its own standard deviations.

\[ z_{x} = \frac{17 - 5}{6} = 2, \qquad z_{y} = \frac{4 - 2}{1} = 2 \]

Both are exactly two standard deviations above their own mean. Relative to their groups they did equally well, and the shaded tails show why that matters: each is beaten by the same 2.28 percent of their own group.

The picture shows the reason in shapes. The tall narrow curve packs its group so tightly around 2 that a gain of 4 is already out in its tail; the low wide curve spreads its group so far that it takes 17 to get the same distance out. Equal tails mean equal standing.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 337 — the z-score allows us to compare data that are scaled differently

39. Example 6.2(c): the same z-score on two scales

Worked example

Suppose X ~ N(5, 6) and Y ~ N(2, 1). The z-score of x equal to 17 was shown to be 2. If y is 4, what is its z-score, and what does the pair of results say?

\[ X \sim N(5, 6), \quad Y \sim N(2, 1) \]

Recall the z-score of x = 17

Why: from Example 6.1.

\[ z_{x} = \frac{17 - 5}{6} = \frac{12}{6} = 2 \]

Substitute y = 4 into Y's own formula

Why: Y has its own mean 2 and standard deviation 1.

\[ z_{y} = \frac{4 - 2}{1} \]

Subtract and divide

Why: 2 units, each worth one standard deviation.

\[ z_{y} = \frac{2}{1} = 2 \]

So 17 and 4 are both two of their own standard deviations to the right of their respective means. As weight gains, they represent the same standardised gain relative to their groups, even though one is more than four times the other in pounds.

The book describes the setting: X is the weight gain of one group trying to gain weight over six weeks and Y the gain of a second group, with a negative gain meaning a loss. Seventeen pounds in the first group and four in the second are the same standardised gain relative to their means.

The one real mistake available here is using the wrong distribution's numbers: putting 4 into X's formula gives minus 0.17 and says the second person did worse than average in a group they do not belong to.

Check by rebuilding each value from z = 2

Why: each distribution's own μ plus 2σ.

\[ 5 + (2)(6) = 17, \qquad 2 + (2)(1) = 4 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 337 — Example 6.2(c)

40. Example 6.4: two generations of Chilean boys

Worked example

From 1984 to 1985, heights of 15 to 18-year-old Chilean males were Y ~ N(172.36, 6.34). From 2009 to 2010 they were X ~ N(170, 6.28). Find the z-scores of x equal to 160.58 cm and y equal to 162.85 cm, interpret each, and compare.

\[ X \sim N(170, 6.28), \qquad Y \sim N(172.36, 6.34) \]

Substitute x = 160.58 into X's formula

Why: the 2009 boy against the 2009 curve.

\[ z_{x} = \frac{160.58 - 170}{6.28} \]

Subtract and divide

Why: 9.42 cm below, and 9.42 over 6.28 is exactly one and a half.

\[ z_{x} = \frac{-9.42}{6.28} = -1.5 \]

Substitute y = 162.85 into Y's formula

Why: the 1984 boy against the 1984 curve.

\[ z_{y} = \frac{162.85 - 172.36}{6.34} \]

Subtract and divide

Why: 9.51 cm below, and 9.51 over 6.34 is also one and a half.

\[ z_{y} = \frac{-9.51}{6.34} = -1.5 \]

Figure (svg): Two nearly identical bell curves, a solid blue one centred at 170 and a dashed orange one centred at 172.36. The blue left tail below 160.58 and the orange left tail below 162.85 are shaded; each boundary is marked with a line of its own colour, and a note says both z equal minus 1.5.

The two boys are different heights, 2.27 cm apart, but each stands 1.5 of his own era's standard deviations below his own era's mean, so each has the same share of his generation below him.

Before interpreting, notice what changed between the eras: the mean fell by 2.36 cm and the spread barely moved. On the picture the two curves almost coincide, which is why raw heights from the two eras look comparable. They are close, but not the same, and the z-scores take care of the difference exactly.

Both boys are 1.5 standard deviations below their own generation's mean. They deviate the same number of standard deviations from their respective means, in the same direction, so each is just as short, relative to his peers, as the other. The book's printed solution writes the first value as minus 160.58; it is a typo for 160.58, and the z-score of minus 1.5 is right.

Check by rebuilding both heights from z = −1.5

Why: each era's own μ plus zσ.

\[ 170 + (-1.5)(6.28) = 160.58, \quad 172.36 + (-1.5)(6.34) = 162.85 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 339 — Example 6.4

41. Try It 6.4: two SAT verbal scores

Worked example

In 2012, 1,664,479 students took the SAT. Verbal scores had mean 496 and standard deviation 114, so X ~ N(496, 114). Find and interpret the z-scores of 325 and 366.21, and compare them.

\[ X \sim N(496, 114) \]

Substitute the first score

Why: well below the mean.

\[ z_{1} = \frac{325 - 496}{114} \]

Subtract and divide

Why: 171 points below, which is one and a half standard deviations.

\[ z_{1} = \frac{-171}{114} = -1.5 \]

Substitute the second score

Why: also below the mean.

\[ z_{2} = \frac{366.21 - 496}{114} \]

Subtract and divide

Why: node gives minus 1.1385.

\[ z_{2} = \frac{-129.79}{114} \approx -1.14 \]

A score of 325 is 1.5 standard deviations below the mean of 496; a score of 366.21 is about 1.14 standard deviations below it. Both are below average, and 325 is further below. The book says their respective means, but both scores come from the same distribution, so here the z-scores simply order two scores on one test.

Since both came from one distribution, you could have compared the raw scores directly. Z-scores are needed when the scales differ; within one scale they agree with the raw order, as they must.

The 325 has a tidy z-score of exactly minus 1.5 for the same reason Example 6.4's values did: the book chose it as the mean minus one and a half standard deviations. When a z-score comes out that clean, it is a hint that the problem was built backwards from x equals μ plus zσ.

Check the order against the raw scores

Why: the lower score must have the more negative z-score.

\[ 325 < 366.21 \quad \text{and} \quad -1.5 < -1.14 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 339 — Try It 6.4

42. Homework 67(c): SAT 700 against ACT 30

Worked example

In one year SAT math scores had mean 514 and standard deviation 117. ACT math scores are roughly normal with mean 21 and standard deviation 5.3. One person scored 700 on the SAT; another scored 30 on the ACT. Who did better with respect to the test they took?

\[ S \sim N(514, 117), \qquad A \sim N(21, 5.3) \]

Standardise the SAT score

Why: against the SAT distribution.

\[ z_{S} = \frac{700 - 514}{117} \]

Subtract and divide

Why: 186 points above the mean.

\[ z_{S} = \frac{186}{117} \approx 1.59 \]

Standardise the ACT score

Why: against the ACT distribution.

\[ z_{A} = \frac{30 - 21}{5.3} \]

Subtract and divide

Why: 9 points above the mean.

\[ z_{A} = \frac{9}{5.3} \approx 1.70 \]

Compare the z-scores

Why: the larger z is further above its own mean.

\[ 1.70 > 1.59 \]

Figure (svg): Two side-by-side bell curves. Left, SAT math with mean 514 and standard deviation 117, ticks at 280, 514 and 748 with z values minus 2, 0 and 2; a blue line at 700 with the tail beyond it shaded, titled z equals 1.59, 5.6 percent score higher. Right, ACT math with mean 21 and standard deviation 5.3, ticks at 10.4, 21 and 31.6; an orange line at 30 with the tail shaded, titled z equals 1.70, 4.5 percent score higher.

The ACT tail is the thinner one. Fewer ACT takers beat a 30 than SAT takers beat a 700, so the 30 is the stronger result.

The ACT taker did better relative to the test they took: a 30 is about 1.70 standard deviations above the ACT mean, while a 700 is about 1.59 above the SAT mean. The raw numbers, 700 and 30, were never comparable. The tail areas in the picture, found with the standard normal in Section 6.2, say the same thing in percentages.

Check with the tail areas

Why: a larger z-score leaves a smaller share of takers above it.

\[ P(Z > 1.70) \approx 0.045 < P(Z > 1.59) \approx 0.056 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 homework, p. 356 — Exercise 67(c)

43. Rank four performances by z-score

Ranking

Four results from four different scales, all from this section or its homework. Put them in order from most below their own mean to most above it. Estimate each z-score roughly first; you need only the order.

Put in order

Order from lowest z-score to highest.

  1. a 7.9 kg, 80 cm girl, weights N(10.2, 0.8)
  2. a 10-point game for Jerome, N(16, 4)
  3. a 168 cm boy, heights N(170, 6.28)
  4. a 700 on SAT math, N(514, 117)
  5. a 30 on ACT math, N(21, 5.3)

Why: The z-scores are about minus 2.88 for the girl, minus 1.5 for Jerome, minus 0.32 for the boy, 1.59 for the SAT score and 1.70 for the ACT score. Only the z-scores can be ordered; the raw values are in kilograms, centimetres, points and scaled scores.

\[ -2.88 < -1.5 < -0.32 < 1.59 < 1.70 \]

Try ordering the raw values and the list is meaningless: 7.9, 10, 30, 168 and 700 are in five different units. The z-scores strip the units away and leave only position, which is what can be compared.

The close pair is the SAT and ACT at the top, already settled in the homework example. The girl at the bottom is the one a doctor would notice, nearly three standard deviations below her reference group.

44. Fill in the missing z-scores and values

Comparison

Each row is a distribution from this section. Some rows give a value and want its z-score; others give a z-score and want the value. Round to two decimal places.

Comparison matrix

distributionxzwhich formula
N(2, 6)203x = μ + zσ
N(12, 6)2-1.67z = (x − μ)/σ
N(9, 5)6.5−0.5x = μ + zσ
N(6, 1.5)5.5-0.33z = (x − μ)/σ
N(8, 9)1.97−0.67x = μ + zσ

These are the book's practice Exercises 15, 23, 17, 25 and 21. For the rows that want x: 2 plus 3 times 6 is 20, 9 minus half of 5 is 6.5, and 8 minus 0.67 times 9 is 1.97. For the rows that want z: minus 10 over 6 is about minus 1.67, and minus 0.5 over 1.5 is about minus 0.33.

Look at the last column before you compute anything. It is decided entirely by which of x and z the row already gives you, which is the same step-zero choice you made for Try It 6.3.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 practice, p. 352 — Exercises 15, 17, 21, 23 and 25

45. Trap: the bigger raw gap wins

Trap

The trap

Homework 67(c) again. A first attempt compares the distances above the means:

\[ 700 - 514 = 186 \]

\[ 30 - 21 = 9 \]

and concludes that the SAT taker did far better, since 186 is much bigger than 9. Wrong.

The fix

A distance means nothing until you know the size of a typical distance on that scale, which is σ. Divide each gap by its own standard deviation.

\[ \frac{186}{117} \approx 1.59 \]

\[ \frac{9}{5.3} \approx 1.70 \]

The ACT taker is further above the mean, in the only units the two tests share. The SAT scale is simply larger, so every gap on it looks bigger. Subtracting the mean is only half of a z-score; the division is the half that makes scales comparable.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 homework, p. 356 — Exercise 67(c)

46. The Empirical Rule

Section

Part 5

47. 68, 95 and 99.7 percent

Concept

Because every normal curve is the same curve in z-units, the share of the area within any fixed number of standard deviations of the mean is the same for all of them. The Empirical Rule records the three shares you will use most.

Figure (svg): A standard bell curve with ticks at mu minus 3 sigma through mu plus 3 sigma. Three nested shaded bands, darkest in the middle, cover one, two and three standard deviations either side. Below the axis, three nested brackets are labelled about 68 percent, about 95 percent and about 99.7 percent.

The brackets nest: each band contains the one inside it. Almost the whole area, 99.7 percent, lies within three standard deviations.

The last line says that almost all the values lie within three standard deviations of the mean, which is why a z-score beyond 3, like the NBA player's 3.5, is so hard to believe. The rule is also called the 68-95-99.7 rule, and it applies to normal distributions: Part 5's last slides test what happens without the bell shape.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 338 — the Empirical Rule and Figure 6.3

48. Decoding a band: μ ± kσ

Notation

Every statement of the rule has the same shape: a percentage, and a band described by the mean and a whole number of standard deviations. Decode the two-standard-deviation line.

Annotate

On: \( P(\mu - 2\sigma < X < \mu + 2\sigma) \approx 0.95 \)

  • The lower edge of the band: two standard deviations below the mean. Its z-score is minus 2.
  • The upper edge: two standard deviations above the mean, z-score 2. The band is symmetric about μ.
  • One randomly chosen value of the variable, which lands inside the band or outside it.
  • About 95 percent of the values, or a probability of about 0.95 for one random value. The exact figure is 0.9545.

Standardise the whole band and the μ and σ disappear, leaving the same statement for the standard normal. That is why the percentages do not depend on which normal distribution you are looking at.

\[ P(-2 < Z < 2) \approx 0.95 \]

Read the percentage two ways: as the share of all values in the band, or as the chance that one randomly chosen value lands in it. For a continuous distribution the two readings are the same number.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 338 — the Empirical Rule in z-scores

49. Why the rule is the same for every normal, and how exact it is

Concept

Take any normal X ~ N(μ, σ). A value lies within two standard deviations of the mean exactly when its z-score lies between minus 2 and 2. Standardising turns one statement into the other, line by line.

\[ \mu - 2\sigma < x < \mu + 2\sigma \]

\[ -2\sigma < x - \mu < 2\sigma \]

\[ -2 < \frac{x - \mu}{\sigma} < 2 \]

The last line mentions no μ and no σ: it is a statement about the standard normal curve alone, so it has one answer for every normal distribution. Computed from the standard normal curve in node, the three answers are these.

bandz-scoresEmpirical Ruleexact
μ ± σ−1 to 1about 68%68.27%
μ ± 2σ−2 to 2about 95%95.45%
μ ± 3σ−3 to 3about 99.7%99.73%

The book's own practice exercises use the exact figures: Exercise 34 asks between which values 68.27 percent of the data lie. Use the rounded rule for quick reasoning, and remember it is a rounding when an answer needs more precision.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 338; practice p. 352 — the Empirical Rule; Exercises 34-35

50. Example 6.5: the bands of N(50, 6)

Worked example

Suppose x has a normal distribution with mean 50 and standard deviation 6. Find the values between which about 68, 95 and 99.7 percent of the x values lie, and their z-scores.

\[ X \sim N(50, 6) \]

One standard deviation

Why: plus and minus one times 6 from the mean.

\[ 50 - 6 = 44, \qquad 50 + 6 = 56 \]

Two standard deviations

Why: plus and minus 12.

\[ 50 - 12 = 38, \qquad 50 + 12 = 62 \]

Three standard deviations

Why: plus and minus 18.

\[ 50 - 18 = 32, \qquad 50 + 18 = 68 \]

Figure (svg): A bell curve for mean 50 and standard deviation 6 with x ticks 32, 38, 44, 50, 56, 62, 68 and z values minus 3 to 3 underneath. Nested shaded bands; below, brackets labelled 68 percent 44 to 56, 95 percent 38 to 62 and 99.7 percent 32 to 68.

Each band edge is the mean plus or minus a whole number of sixes. Read the z row and you see the same −1 to 1, −2 to 2 and −3 to 3 as on every other normal curve.

So about 68 percent of the values lie between 44 and 56, about 95 percent between 38 and 62, and about 99.7 percent between 32 and 68. The z-scores of the band edges are minus 1 and 1, minus 2 and 2, and minus 3 and 3.

The procedure is x equals μ plus zσ used six times, with z equal to minus 3 through 3. That is all the Empirical Rule asks of you in original units: build the band edges from the mean and the standard deviation.

Look at the picture's two rows of labels. The top row changes from problem to problem; the bottom row, minus 3 to 3, never does. That is the Empirical Rule in one image: the bands are fixed on the z row, and only their translation into units depends on μ and σ.

Check a band edge by computing its z-score

Why: 38 should sit exactly two standard deviations below.

\[ \frac{38 - 50}{6} = \frac{-12}{6} = -2 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 339 — Example 6.5

51. Try It 6.5: the middle 68 percent of N(25, 5)

Fill the middle

Suppose X has a normal distribution with mean 25 and standard deviation five. Between what values of x do about 68 percent of the values lie? Fill in the two band edges.

Fill in the blanks

25 - 5 = 20, \qquad 25 + 5 = 30

Why: Sixty-eight percent is the one-standard-deviation band, so step one standard deviation of 5 each way from the mean of 25: from 20 to 30.

About 68 percent of the values lie between 20 and 30. The two edges have z-scores minus 1 and 1, as every 68 percent band does.

The common slip here is to use the wrong band, stepping two standard deviations and getting 15 to 35. That is the 95 percent band. Match the percentage to the number of standard deviations before you compute: 68 is one, 95 is two, 99.7 is three.

\[ P(20 < X < 30) \approx 0.68 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, pp. 339-340 — Try It 6.5

52. Example 6.6: the bands for 1984 Chilean heights

Worked example

From 1984 to 1985, heights of 15 to 18-year-old Chilean males were Y ~ N(172.36, 6.34). Between what two values do about 68, 95 and 99.7 percent of the heights lie?

\[ Y \sim N(172.36, 6.34) \]

Multiply σ by 1, 2 and 3

Why: the half-widths of the three bands.

\[ 6.34, \quad 12.68, \quad 19.02 \]

68 percent: μ ± 6.34

Why: subtract and add the first half-width.

\[ 172.36 - 6.34 = 166.02, \quad 172.36 + 6.34 = 178.70 \]

95 percent: μ ± 12.68

Why: the second half-width.

\[ 172.36 - 12.68 = 159.68, \quad 172.36 + 12.68 = 185.04 \]

99.7 percent: μ ± 19.02

Why: the third.

\[ 172.36 - 19.02 = 153.34, \quad 172.36 + 19.02 = 191.38 \]

So about 68 percent of these boys were between 166.02 cm and 178.70 cm tall, about 95 percent between 159.68 and 185.04 cm, and nearly all, 99.7 percent, between 153.34 and 191.38 cm. The z-scores are minus 1 and 1, minus 2 and 2, minus 3 and 3. The book prints 178.7 for 178.70.

Compare the boy of Example 6.4, 162.85 cm: he lies inside the 95 percent band but outside the 68 percent band, which is exactly what a z-score of minus 1.5 means.

Check that each band is centred on the mean

Why: the midpoint of each pair of edges must be 172.36.

\[ \frac{166.02 + 178.70}{2} = \frac{159.68 + 185.04}{2} = \frac{153.34 + 191.38}{2} = 172.36 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 340 — Example 6.6

53. Try It 6.6: a college entrance exam

Worked example

Scores on a college entrance exam are approximately normal with mean 52 points and standard deviation 11 points. Find the three Empirical Rule bands and their z-scores.

\[ X \sim N(52, 11) \]

68 percent: one standard deviation each way

Why: 52 plus and minus 11.

\[ 52 - 11 = 41, \quad 52 + 11 = 63 \]

95 percent: two standard deviations

Why: 52 plus and minus 22.

\[ 52 - 22 = 30, \quad 52 + 22 = 74 \]

99.7 percent: three standard deviations

Why: 52 plus and minus 33.

\[ 52 - 33 = 19, \quad 52 + 33 = 85 \]

About 68 percent of test takers score between 41 and 63, about 95 percent between 30 and 74, and about 99.7 percent between 19 and 85. The z-scores of the edges are minus 1 and 1, minus 2 and 2, and minus 3 and 3.

The bottom edge of the widest band, 19, is worth a moment of thought. If the exam's minimum possible score were, say, 20, the model would be predicting a few impossible scores, which is a reminder that approximately normal means the model fits the middle well and the extremes only roughly.

Notice the widths: 22, 44 and 66 points. Each band is twice, four times and six times σ wide, so the three widths are in the ratio 1 to 2 to 3, whatever the distribution. A student who gets band widths out of that ratio has mixed up a step.

Check the widths

Why: each band is 2kσ wide for k equal to 1, 2 and 3.

\[ 63 - 41 = 22 = 2(11), \quad 74 - 30 = 44 = 4(11), \quad 85 - 19 = 66 = 6(11) \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 340 — Try It 6.6

54. Halving and subtracting: the pieces of the curve

Picture it

The Empirical Rule gives three symmetric bands, but questions often ask about one side, or about the ring between two bands. Two moves answer all of them: halve a band using symmetry, and subtract one band from a larger one.

Figure (svg): A standard bell curve cut at z equals minus 3, minus 2, minus 1, 0, 1, 2 and 3 into eight coloured pieces. The two central pieces are labelled 34 percent, the next pair 13.5 percent, the next pair 2.35 percent, and the two outermost tails 0.15 percent.

Symmetry halves each band, and subtracting one band from the next gives the rings between them. The eight pieces add to 100 percent.

\[ \text{mean to } +1\sigma: \quad \frac{68}{2} = 34\% \]

\[ +1\sigma \text{ to } +2\sigma: \quad \frac{95 - 68}{2} = 13.5\% \]

\[ +2\sigma \text{ to } +3\sigma: \quad \frac{99.7 - 95}{2} = 2.35\% \]

\[ \text{beyond } +3\sigma: \quad \frac{100 - 99.7}{2} = 0.15\% \]

Always draw the curve and shade the piece first, then decide which bands to subtract and whether to halve. The eight pieces in the picture add to 100 percent, which is a quick check that you have not double-counted one.

The pieces also combine. Below the mean plus one standard deviation is the whole left half, 50 percent, plus the 34 percent piece: about 84 percent. Above the mean minus two standard deviations is 13.5 plus 34 plus 50, about 97.5 percent. Each is just a sum of pieces you can see.

\[ P(X < \mu + \sigma) \approx 50\% + 34\% = 84\% \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 practice, p. 352 — Exercises 33, 36-38

55. Exercises 33 and 37: rings and half-bands

Worked example

Practice Exercise 33 asks what percent of x values lie between the second and third standard deviations, on both sides. Exercise 37 asks what percent lie between the mean and three standard deviations, on one side.

Exercise 33: subtract the 95 band from the 99.7 band

Why: what is left is the two rings, one on each side.

\[ 99.7\% - 95\% = 4.7\% \]

Recompute with the exact band areas

Why: the rounded rule overstates this small difference.

\[ 99.73\% - 95.45\% = 4.28\% \]

The book's answer is about 4 percent, which matches the exact 4.28. Subtracting two rounded percentages loses precision when the difference is small, which is why the rule gives 4.7 here. Either is a fair answer to an about question, but know where each comes from.

Exercise 37: halve the three-standard-deviation band

Why: the mean to 3σ is one side of it.

\[ \frac{99.7\%}{2} \approx 49.85\% \]

So about 4 percent of values lie between two and three standard deviations from the mean, counting both sides, and about 49.85 percent lie between the mean and three standard deviations above it.

Check against the pieces

Why: the one-sided half-bands must add to one half.

\[ 34 + 13.5 + 2.35 = 49.85 \quad \text{and} \quad 49.85 + 0.15 = 50 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 practice, p. 352 — Exercises 33 and 37

56. What unusual means: beyond two standard deviations

Concept

The Empirical Rule gives you a ready-made yardstick for how surprising a value is. Within one standard deviation of the mean is the ordinary middle, where about two values in three land. Between one and two standard deviations is less common but routine. Beyond two is where about one value in twenty lands, split between the two tails.

\[ |z| > 2: \quad \text{about } 5\% \text{ of values} \]

\[ |z| > 3: \quad \text{about } 0.3\% \text{ of values} \]

So a z-score tells you at a glance how unusual a value is, whatever the units. The 168 cm boy, at minus 0.32, is in the ordinary middle. Jerome's ten-point game, at minus 1.5, is a below-par night but not rare. The 7.9 kg girl, at minus 2.88, is in the outer five percent, and the claimed NBA height, at 3.5, is beyond even the 99.7 percent band.

z-scorewhere it sitsroughly how often, both tails together
between −1 and 1the ordinary middleabout 68% of values
between 1 and 2 in sizeuncommon, but routineabout 27% of values
between 2 and 3 in sizeunusualabout 4.7% of values
beyond 3 in sizevery unusualabout 0.3% of values

Many courses and many clinics use two standard deviations as the line for unusual, as the growth charts did. It is a convention, not a law, but it is built directly on the 95 percent band, and you will see the same cut-off, 1.96 to be exact, when Chapter 8 builds confidence intervals.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 338 — the Empirical Rule

57. Homework 64: a friend's blood pressure

Worked example

Male systolic blood pressure, in millimetres, is normal with mean 125 and standard deviation 14. Part (a): find the z-scores of 100 and 150. Part (b): a friend says his blood pressure is 2.5 standard deviations below the mean, but also that it is between 100 and 150. What would you say to him?

\[ X \sim N(125, 14) \]

(a) Standardise 100

Why: 25 below the mean.

\[ z = \frac{100 - 125}{14} = \frac{-25}{14} \approx -1.79 \]

Standardise 150

Why: 25 above the mean.

\[ z = \frac{150 - 125}{14} = \frac{25}{14} \approx 1.79 \]

(b) Convert the friend's z-score to a pressure

Why: x equals μ plus zσ with z of minus 2.5.

\[ x = 125 + (-2.5)(14) = 125 - 35 = 90 \]

Figure (svg): A bell curve of male systolic blood pressure with mean 125 and standard deviation 14, ticks from 83 to 167 one standard deviation apart and z values minus 3 to 3 beneath. The band from 100 to 150 is shaded blue, labelled z from minus 1.79 to 1.79. A red line at 90, z equals minus 2.5, stands outside the band on the left.

The friend's two claims cannot both be true: 2.5 standard deviations below 125 is 90, well outside the band from 100 to 150.

Part (a) is symmetric on purpose: 100 and 150 are each 25 millimetres from the mean, so their z-scores are equal in size and opposite in sign. The band from 100 to 150 is therefore the band within 1.79 standard deviations of the mean.

His two statements contradict each other. A pressure between 100 and 150 has a z-score between minus 1.79 and 1.79, but 2.5 standard deviations below the mean is 90, which is outside that range. At least one of his claims is wrong, and you would tell him so.

The Empirical Rule adds context: 100 to 150 is a bit less than two standard deviations either side, so most men, a little under 95 percent, fall in it. The friend's first claim would put him in the lowest one percent or so.

Check the contradiction in z-units

Why: minus 2.5 must lie inside the band's z-range for both claims to hold, and it does not.

\[ -2.5 < -1.79 \]

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 homework, p. 356 — Exercise 64

58. How much lies more than two standard deviations above the mean?

Prediction

Exam scores are normal. What share of students score more than two standard deviations above the mean? Use only the Empirical Rule and symmetry.

Predict first

About what percent of a normal distribution lies more than two standard deviations above the mean?

  • about 2.5 percent
  • about 5 percent
  • about 95 percent
  • about 13.5 percent

Correct: about 2.5 percent

Why: About 95 percent lies within two standard deviations, so about 5 percent lies outside, split equally between the two tails by symmetry. Above the upper edge is half of that, about 2.5 percent.

\[ \frac{100\% - 95\%}{2} = 2.5\% \]

The five percent option forgets to halve: it counts both tails, the very low and the very high. The 13.5 percent option is the ring between one and two standard deviations, a different piece. The exact value, from the standard normal curve, is 2.28 percent, the tail shaded in the weight-gain picture of Part 4.

This single number is used constantly later in the course: a result more than two standard deviations out, on one side, happens only about one time in forty.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 338 — the Empirical Rule

59. Does the Empirical Rule work for any distribution?

Edge cases

The Empirical Rule came from the shape of the normal curve. Push it outside that shape. Waiting times often follow an exponential distribution, from Chapter 5, which is heavily skewed. With mean 1 minute, its standard deviation is also 1 minute.

Figure (svg): A decreasing exponential density starting at height 1 at x equals 0 and falling towards the axis. The region from 0 to 2, which is the mean plus or minus one standard deviation clipped at zero, is shaded and labelled 86.5 percent, not 68.

This curve is not a bell. The band of one standard deviation either side of the mean holds 86.5 percent of the area, and the left half of the band is cut off at zero.

Discussion prompt

For this exponential distribution with mean 1 and standard deviation 1, the band within one standard deviation of the mean runs from 0 to 2. The area over that band is 1 minus e to the minus 2. Is that close to 68 percent? What does this say about when you may use the Empirical Rule?

This is the book's warning that the normal is widely used and even more widely abused, made concrete. Before quoting 68, 95 or 99.7 percent, look at a histogram of the data. If it is skewed, like waiting times or incomes, the rule gives the wrong answer with complete confidence.

The homework's Exercise 62 makes a related point: a quantity with a normal shape cannot also be exponential or uniform. The shape comes first, and the rule comes with it.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, p. 335 — the normal cannot be applied to everything

60. Putting it together

Section

Part 6

61. Pattern: one relationship, two directions

Pattern

Every problem in this section ran through the same short procedure. Only the direction changed.

Figure (svg): A flow diagram. Read mu and sigma from X distributed N of mu and sigma. Then branch: a value to a z-score by z equals x minus mu over sigma, or a z-score to a value by x equals mu plus z sigma. Both lead to: say it in words, how many standard deviations and which side. Then either compare with other scales by z, or read a percentage off the Empirical Rule.

Two directions, one formula. Whichever way you go, the answer is not finished until it is a sentence about standard deviations from the mean, in context.
  1. Read μ and σ from X ~ N(μ, σ), remembering the second number is the standard deviation.
  2. Decide the direction: given a value, find z; given a z-score or a number of standard deviations, find x.
  3. Predict the side of the mean before computing, so the sign of your answer has something to agree with.
  4. Compute with z = (x − μ)/σ or x = μ + zσ, one move per line.
  5. Say it in context: so many standard deviations above or below the mean, in the units of the problem.
  6. Then use it: compare scales by z-scores, or read a share off the 68-95-99.7 bands.

\[ z = \frac{x - \mu}{\sigma} \quad \Longleftrightarrow \quad x = \mu + z\sigma \]

The arithmetic is never hard in this section; the reasoning is where the marks go. Every common error you have seen, the swapped subtraction, the missing division, the dropped sign, the wrong band, is caught by the prediction in the third move and the sentence in the fifth.

62. Check: a recovery time

Check

Homework Exercises 60 and 61. Standardise, and do the direction check before you pick.

Check your understanding

Recovery time from a surgical procedure is normal with mean 5.3 days and standard deviation 2.1 days. What is the z-score for a patient who takes ten days to recover?

  • A. about 2.2 (correct)
  • B. about 4.8
  • C. about 0.9
  • D. about −2.2

Answer: A

Why: Ten days is 4.7 days above the mean of 5.3, and 4.7 divided by the standard deviation 2.1 is about 2.24. Ten is above the mean, so the z-score is positive: about 2.2 standard deviations above.

Why B tempts people
That is 10 divided by 2.1, which forgets to subtract the mean first. A z-score measures distance from the mean, not from zero.
Why C tempts people
That divides the distance 4.7 by the mean 5.3 instead of the standard deviation, giving a proportion rather than a count of standard deviations.
Why D tempts people
That subtracts the wrong way round, the mean minus the value. Ten days is longer than average, so the sign must be positive.

Every wrong option is a slip you saw in Part 2: forgetting the mean, dividing by the wrong number, and flipping the sign. The direction check alone rules out the negative option in a second.

Exercise 60, the partner question, asks for the median recovery time. The curve is symmetric, so the median equals the mean: 5.3 days.

A z-score of about 2.2 also means this patient is in the slowest few percent: past two standard deviations, where only about 2.5 percent of a normal distribution lies.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 homework, p. 355 — Exercises 60-61

63. Check: reading Kyle's z-score

Check

Homework Exercise 65. Read the z-score as a sentence before touching the arithmetic.

Check your understanding

Male systolic blood pressure is normal with mean 125 and standard deviation 14. Kyle's doctor tells him his z-score is 1.75. Which statement is correct?

  • A. Kyle's pressure is 1.75 standard deviations above the mean, which is 149.5 (correct)
  • B. Kyle's pressure is 1.75 times the average, which is 218.75
  • C. Kyle's pressure is 1.75 above the average, which is 126.75
  • D. Kyle's pressure is 175

Answer: A

Why: A z-score counts standard deviations from the mean. Kyle is 1.75 standard deviations of 14 above 125, which is 24.5 above, so his systolic pressure is 149.5.

Why B tempts people
A z-score is never a multiple of the mean. It compares a distance from the mean with the standard deviation.
Why C tempts people
That reads 1.75 as millimetres. The z-score is in standard deviations, and each one is 14 millimetres.
Why D tempts people
That treats the z-score as the reading itself, scaled up by 100. The z-score has no units; it has to be converted back with the mean and standard deviation.

\[ x = 125 + (1.75)(14) = 125 + 24.5 = 149.5 \]

The book lists all four readings and asks which are correct; only the one in standard deviations is. The other three each forget one piece of the definition: the mean, the division by σ, or both.

A pressure of 149.5 is high but inside the 95 percent band, which runs from 97 to 153. It is worth watching, not alarming.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 homework, p. 356 — Exercise 65

64. Check: an Empirical Rule band

Check

Practice Exercises 34 and 36. Match the percentage to the band first.

Check your understanding

Suppose X ~ N(15, 3). Between which two values do about 68 percent of the data lie, centred at the mean?

  • A. 12 and 18 (correct)
  • B. 9 and 21
  • C. 6 and 24
  • D. 15 and 18

Answer: A

Why: Sixty-eight percent is the band within one standard deviation of the mean, so step 3 each way from 15: from 12 to 18. The book states the figure as 68.27 percent, the exact share.

Why B tempts people
That steps two standard deviations each way, which is the 95 percent band.
Why C tempts people
That steps three standard deviations each way, which is the 99.7 percent band.
Why D tempts people
That is only the upper half of the band, from the mean to one standard deviation above. It holds about 34 percent, half of 68.

The last distractor is Exercise 36's question in disguise: for X ~ N(minus 3, 1), about 34 percent of the data lie between minus 3 and minus 2, or between minus 4 and minus 3. Half of a symmetric band is a legitimate question; it just is not this one.

Match percentage to band before computing: 68 is one standard deviation, 95 is two, 99.7 is three.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 practice, p. 352 — Exercises 34 and 36

65. Explain why one curve is enough

Explain it to yourself

There are infinitely many normal distributions, yet this section reduced them all to one curve. Explaining how is the best test of whether the section has landed.

Discussion prompt

In a few sentences, explain why the Empirical Rule's percentages are the same for every normal distribution, using what a z-score measures. Use the two eras of Chilean heights as your example.

A strong answer names the two moves, subtracting the mean and dividing by σ, and says what each does to the curve. It also notices the condition hidden underneath: all of this relies on the distribution being normal in the first place, which is why the exponential waiting times broke the rule.

If your explanation used the word shape, that is the key: the mean and standard deviation set the position and the scale, and standardising removes both, leaving only the shape, which every normal distribution shares.

66. Exit ticket: SAT math

Exit ticket

Homework Exercise 67(a)(b). SAT math scores one year were normal with mean 520 and standard deviation 115.

Discussion prompt

Calculate the z-score for an SAT math score of 720 and interpret it in a complete sentence. Then find the score that is 1.5 standard deviations above the mean, and use the Empirical Rule to say roughly what share of students scored higher than 720.

The first mark is the sign and the division; the second is choosing x equals μ plus zσ for the reverse direction. The last sentence earns the rest: the Empirical Rule cannot give an exact share for 1.74, but the pieces bracket it, beyond two standard deviations 2.5 percent and beyond one 16 percent.

Being able to say between 2.5 and 16 percent, with reasons, is exactly the kind of estimate the next section sharpens into a precise probability.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1 homework, p. 356 — Exercise 67(a)(b)

67. Recap

Recap

A normal distribution is a symmetric bell curve fixed by two numbers: the mean μ, which says where it sits, and the standard deviation σ, which says how wide it is. The area under it is always one, so a wider curve is a lower one.

ideaformularead it as
notationX ~ N(μ, σ)normal, mean μ, standard deviation σ (never the variance)
z-scorez = (x − μ)/σhow many standard deviations x is above (+) or below (−) the mean
back to a valuex = μ + zσstart at the mean, walk z standard deviations
standard normalZ ~ N(0, 1)the one curve every normal becomes when standardised
comparing scalescompare z, not xposition relative to each distribution's own mean and spread
Empirical Rule68 – 95 – 99.7within 1, 2, 3 standard deviations; exact 68.27, 95.45, 99.73

The work is in the direction and the sentence: decide whether you are going from a value to a z-score or back, predict the side of the mean, and finish with how many standard deviations above or below the mean, in context. The Empirical Rule then turns a z-score into a rough share, but only for data that really are bell-shaped.

The Introductory Business Statistics text covers the same material in its own §6.1, with business contexts, if you want more practice. Section 6.2 uses the standard normal curve to turn any z-score into an exact probability.

OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution §6.1, pp. 335-340 — the whole section

OpenStax Introductory Business Statistics 2e, §6.1 The Standard Normal Distribution §6.1 The Standard Normal Distribution — the same ideas in business contexts

Sources

  1. OpenStax Introductory Statistics 2e, §6.1 The Standard Normal Distribution — Illowsky & Dean, OpenStax / Rice University, CC BY 4.0, pp. 335-340
  2. OpenStax Introductory Business Statistics 2e, §6.1 The Standard Normal 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