One-sample t-test

One-sample t-test

Spec SG1 · AQA A-Level Further Maths: Revision notes

Studying this course? Keep your progress in one place.

Key facts

  • Use a t-test for the mean of a normal population when the variance is unknown and the sample is small.
  • The test statistic is t=xˉ−μ0s/nt=\dfrac{\bar x-\mu_0}{s/\sqrt n}, with ν=n−1\nu=n-1 degrees of freedom.
  • Use the unbiased estimate s2=1n−1(∑x2−(∑x)2n)s^2=\dfrac{1}{n-1}\left(\sum x^2-\dfrac{(\sum x)^2}{n}\right).
  • For a two-tailed test, put half the significance level in each tail.
  • If tt is in the critical region, reject H0H_0; always conclude in context.

When to use a t-test

Use the tt-distribution when you estimate σ\sigma by ss from a small sample of a normal population.

The statistic t=xˉ−μ0s/nt=\dfrac{\bar x-\mu_0}{s/\sqrt n} follows the t-distribution with ν=n−1\nu=n-1 degrees of freedom.

It is symmetric about 00 like the normal but has heavier tails, allowing for the extra uncertainty from estimating σ\sigma. As ν\nu increases it approaches the standard normal.

The test needs a normal population and a random sample.

−4−3−2−112340.10.20.30.4xyt, 3 degrees of freedomstandard normal
Heavier tails: t with 3 degrees of freedom against the standard normal (dashed).

z-test

Population variance:
Known
Distribution used:
Normal
Normal population needed:
For small nn

t-test

Population variance:
Unknown, estimated by ss
Distribution used:
tt with n−1n-1 degrees of freedom
Normal population needed:
Yes

A sample of 1212 values is used in a one-sample tt-test. How many degrees of freedom?

Finding s from the data

Use the unbiased estimate of the variance, which divides by n−1n-1, then take the square root.

The unbiased estimate s2s^2 is the one you need for the test. If the question gives the variance, take the square root to get ss before dividing by n\sqrt n.

Check the wording: an 'unbiased estimate of the variance' is already s2s^2, whereas a 'sample variance' with divisor nn must be multiplied by nn−1\frac{n}{n-1}.

  • Unbiased variances2=1n−1(∑x2−(∑x)2n)s^2=\dfrac{1}{n-1}\left(\sum x^2-\dfrac{(\sum x)^2}{n}\right)
  • Divisor nn to n−1n-1s2=nn−1×s^2=\dfrac{n}{n-1}\times sample variance

Worked example

A sample has n=10n=10, ∑x=1840\sum x=1840 and ∑x2=339 136\sum x^2=339\,136. Find xˉ\bar x and ss.

A sample of n=4n=4 values has variance 3636 calculated with divisor nn. What is the unbiased estimate s2s^2?

Hypotheses and critical regions

State H0H_0 and H1H_1 about μ\mu, then read the critical value from the tt-table using ν=n−1\nu=n-1 and the tail probability.

Test H0:μ=μ0H_0:\mu=\mu_0 against H1:μ>μ0H_1:\mu>\mu_0 or H1:μ<μ0H_1:\mu<\mu_0 (one-tailed), or H1:μ≠μ0H_1:\mu\neq\mu_0 (two-tailed).

The significance level goes wholly in one tail for a one-tailed test, and half in each tail for a two-tailed test. For a 5%5\% two-tailed test use the 2.5%2.5\% column.

The critical region is the set of tt values beyond the critical value.

One-tailed, 5%

Table column:
5%5\%
Critical value for ν=8\nu=8:
1.8601.860

Two-tailed, 5%

Table column:
2.5%2.5\%
Critical value for ν=8\nu=8:
±2.306\pm2.306

For a two-tailed test at the 1%1\% significance level, which table column do you use?

Carrying out the test

Calculate tt, compare it with the critical value, then state the conclusion in context with cautious wording.

If tt is in the critical region, reject H0.

If it is not, say there is insufficient evidence against H0H_0. Do not say that H0H_0 is true.

  1. 1

    Hypotheses

    H0H_0 and H1H_1 about μ\mu

  2. 2

    Statistic

    Find xˉ\bar x, ss, then tt

  3. 3

    Critical value

    Use ν=n−1\nu=n-1 and the tail

  4. 4

    Compare

    Is tt in the critical region?

  5. 5

    Conclude

    In context, cautiously

Worked example

A café claims its coffees hold 250250 ml. A sample of n=9n=9 has xˉ=243.4\bar x=243.4 and s=9s=9. Test at the 5%5\% level whether the mean is below 250250 ml.

A test statistic lies outside the critical region. What is the correct conclusion?

Try an exam question

A teacher claims that the mean reaction time of her class is 0.250.25 s. A random sample of 99 students has mean 0.2740.274 s and unbiased standard deviation s=0.036s=0.036 s. Assuming reaction times are normally distributed, test at the 5%5\% level whether the mean differs from 0.250.25 s. (The two-tailed critical value for ν=8\nu=8 is 2.3062.306.)

[5 marks]

That's the notes covered.

Carry on to the next subtopic.

Try a question

How many degrees of freedom does the tt-statistic have in a test of the mean using this sample?

1 mark, marked against the mark scheme

Revision notes