Lesson 7 of 8 · 16 min

Standardising and z-scores

Convert values to z-scores and back, use the standard normal distribution and its symmetry, and compare results that come from different distributions.

What you will be able to do

  • Calculate a zz-score z=x−μσz = \frac{x - \mu}{\sigma} and interpret its size and sign.
  • Find the value x=μ+zσx = \mu + z\sigma that has a given zz-score.
  • Use P(X<x)=P(Z<z)P(X < x) = P(Z < z) with Z∼N(0,1)Z \sim N(0, 1), and the symmetry of the standard normal curve.
  • Compare performances from different distributions using standardised scores.

Mia scored 72 on a Maths test and 80 on an English test. Which was the better result? The raw marks say English, but that ignores how everyone else went. If the Maths class averaged 65 with standard deviation 5 and the English class averaged 74 with standard deviation 6, the question becomes: on which test was Mia further above average, measured in standard deviations? That measure is a zz-score.

Definition

z-score (standardised score)

For a value xx from a distribution with mean μ\mu and standard deviation σ\sigma,

z=x−μσ.z = \frac{x - \mu}{\sigma}.

zz is the number of standard deviations xx lies from the mean: positive above the mean, negative below, 0 at the mean. Rearranged, x=μ+zσx = \mu + z\sigma.

Reaction times in a study have mean 70 ms and standard deviation 8 ms. Find the zz-score of a reaction time of 82 ms.

Worked example

Which test result was better?

  • Simple familiar
  • Technology free

Mia scored 72 on a Maths test where the class mean was 65 and the standard deviation was 5. She scored 80 on an English test where the mean was 74 and the standard deviation was 6. Relative to her classes, on which test did Mia perform better? Justify your answer.

3 steps

Try it yourself first, then check the working one step at a time.

Method

  1. Calculate z=x−μσz = \frac{x - \mu}{\sigma} for each result.

  2. Higher zz is relatively better (unless low values are better, as in race times).

  3. State the conclusion in context.

Watch out for

  • Comparing raw scores or raw distances above the mean (7 and 6) without dividing by σ\sigma.

Standardising does more than compare. If X∼N(μ,σ2)X \sim N(\mu, \sigma^2), then subtracting μ\mu shifts the curve to centre 0, and dividing by σ\sigma rescales its spread to 1. The result is the standard normal variable

Z=X−μσ∼N(0,1).Z = \frac{X - \mu}{\sigma} \sim N(0, 1).

Every normal probability is a standard normal probability in disguise: P(X<x)=P(Z<x−μσ)P(X < x) = P\left(Z < \frac{x - \mu}{\sigma}\right).

Interactive

The standard normal curve

The standard normal distribution Z∼N(0,1)Z \sim N(0, 1). Move the bound and read off P(Z<z)P(Z < z). Try z=1.5z = 1.5 and z=−1.5z = -1.5: the two areas add to 1, because the curve is symmetric.

The lengths of a species of fish have mean 60 cm and standard deviation 4 cm. What length has a zz-score of −1.25-1.25?

in cm

Key idea

Symmetry of the standard normal curve

For any a>0a > 0:

  • P(Z>a)=1−P(Z<a)P(Z > a) = 1 - P(Z < a) (complement)
  • P(Z<−a)=P(Z>a)P(Z < -a) = P(Z > a) (mirror image)
  • P(−a<Z<a)=2P(Z<a)−1P(-a < Z < a) = 2P(Z < a) - 1

With one given probability you can find many others without technology.

By symmetry, P(Z<−1.2)=P(Z>1.2)P(Z < -1.2) = P(Z > 1.2). The unshaded middle is P(−1.2<Z<1.2)=1−2P(Z>1.2)P(-1.2 < Z < 1.2) = 1 - 2P(Z > 1.2).

Given that P(Z<1.2)=0.8849P(Z < 1.2) = 0.8849, find P(−1.2<Z<1.2)P(-1.2 < Z < 1.2) without technology.

Worked example

Using one given probability and symmetry

  • Complex familiar
  • Technology free

X∼N(50,42)X \sim N(50, 4^2) and P(X<56)=0.9332P(X < 56) = 0.9332. Without technology, find:

(a) P(X>56)P(X > 56) (b) P(X<44)P(X < 44) (c) P(44<X<56)P(44 < X < 56) (d) P(50<X<56)P(50 < X < 56)

5 steps

Try it yourself first, then check the working one step at a time.

Method

  1. Standardise to spot mirror-image values.

  2. Use 1−P1 - P, P(Z<−a)=P(Z>a)P(Z < -a) = P(Z > a) and P(X<μ)=0.5P(X < \mu) = 0.5.

Watch out for

  • Writing P(X<44)=1−0.9332P(X < 44) = 1 - 0.9332 without explaining the symmetry: it is right, but say why.

Common mistake

A z-score is not a probability

z=1.5z = 1.5 means "1.5 standard deviations above the mean". It is not a probability (probabilities never exceed 1, and zz can be negative). To turn zz into a probability, use technology or the 68–95–99.7% rule: P(Z<1.5)≈0.9332P(Z < 1.5) \approx 0.9332.

Exam tip

Comparing with z-scores

In a "compare" or "which is better" question, calculate both zz-scores, compare them, and finish with a sentence in context: "Mia's Maths result was 1.4 standard deviations above the class mean, compared with 1.0 for English, so relative to her classes Maths was the better result." If a lower value is better (a race time), say so: the more negative zz-score is better.

Summary

  • z=x−μσz = \frac{x - \mu}{\sigma} counts standard deviations from the mean; x=μ+zσx = \mu + z\sigma reverses it.

  • If X∼N(μ,σ2)X \sim N(\mu, \sigma^2) then Z=X−μσ∼N(0,1)Z = \frac{X - \mu}{\sigma} \sim N(0, 1), so P(X<x)=P(Z<z)P(X < x) = P(Z < z).

  • Symmetry: P(Z<−a)=P(Z>a)P(Z < -a) = P(Z > a) and P(−a<Z<a)=2P(Z<a)−1P(-a < Z < a) = 2P(Z < a) - 1.

  • Compare results from different distributions by their zz-scores, and interpret in context.

Finished the lesson?

Try the 3 quick checks above and the lesson completes itself, or mark it complete now.

Lesson 7 of 80% read

Save your progress with a free account