Inference using normal and t-distributions
Введено| English | Значение |
|---|---|
| sample/ˈsæmpl/ | выборка |
| distribution/ˌdɪstrɪˈbjuːʃn/ | распределение |
| variance/ˈveərɪəns/ | дисперсия |
| tail/teɪl/ | хвост |
| confidence interval/ˈkɒnfɪdəns ˈɪntəvl/ | доверительный интервал |
| degrees of freedom/dɪˈɡriːz ɒv ˈfriːdəm/ | степени свободы |
| standard error/ˈstændəd ˈerə/ | стандартная ошибка |
| paired/peəd/ | парные данные |
| pooled estimate/puːld ˈestɪmət/ | объединённая оценка |
Small samples 样本, honest uncertainty
- With a huge sample you can use the normal distribution 分布. But with only, say, 8 measurements and an unknown spread, the normal is too confident.
- William Gosset, brewing at Guinness, invented the $t$-distribution to handle exactly this — small samples with unknown variance 方差.
When to use the t-distribution
- Use the $t$-distribution when the sample is small, the population variance is unknown, and the population is normal (state this assumption in answers).
- Estimate the variance with the unbiased estimate $s^2=\dfrac{1}{n-1}\left(\sum x^2-\dfrac{(\sum x)^2}{n}\right)$.
- It is bell-shaped like the normal but with heavier tails 尾部 — wider, to admit the extra uncertainty.

The t-distribution has a lower peak and heavier tails than the normal
Inference & the normal curve
P(Z < z)
Tests and intervals read площади under the normal curve — slide z to find them.
You use the t-distribution (rather than the normal) when:
Сложность 1/5 · 100 монет + небольшой бонус за первое завершение сегодня
Small n with unknown σ → estimate with s and use t.
The t-distribution has lighter tails than the normal distribution.
Сложность 1/5 · 100 монет + небольшой бонус за первое завершение сегодня
False. The t-distribution has heavier tails, reflecting the extra uncertainty from estimating the variance; it approaches the normal as the degrees of freedom grow.
A sample has n = 5, Σx = 20 and Σx² = 90. Find the unbiased estimate of the population variance.
Сложность 2/5 · 140 монет + небольшой бонус за первое завершение сегодня
s² = (90 − 20²/5) ÷ 4 = (90 − 80) ÷ 4 = 2.5.
Confidence interval 置信区间 for the mean
- A confidence interval for the population mean:
- $t$ comes from the $t$-tables with $n - 1$ degrees of freedom 自由度; $\dfrac{s}{\sqrt n}$ is the standard error 标准误.
For a single sample of size n = 10, how many degrees of freedom does the t-value use?
Сложность 2/5 · 140 монет + небольшой бонус за первое завершение сегодня
Degrees of freedom = n − 1 = 9.
The quantity s/√n is called the standard ______.
Сложность 1/5 · 100 монет + небольшой бонус за первое завершение сегодня
The standard error of the mean is s/√n.
A single sample has size 16. How many degrees of freedom does the t-value use, and why is that not 16?
One degree of freedom is used in estimating the sample standard deviation, leaving n minus 1.
Worked example
- A sample of $n = 10$ has $s = 4$; for 95% with 9 degrees of freedom, $t = 2.262$.
- Margin $= t\dfrac{s}{\sqrt n} = 2.262 \times \dfrac{4}{\sqrt{10}} = 2.86$ (2 dp).
- If $\bar x=50.0$, the 95% interval is $50.0\pm2.86$, which is $(47.14,\ 52.86)$. Interpret it: if we took many samples, about 95% of the intervals made this way would contain $\mu$.
A sample n = 10 has s = 4. With t = 2.262, what is the margin t × s/√n? (2 dp)
Сложность 2/5 · 140 монет + небольшой бонус за первое завершение сегодня
2.262 × 4/√10 = 2.262 × 1.265 = 2.86.
Comparing two populations
- Paired data (the same people measured twice): work with the differences $d$. Test $\bar d$ with a one-sample $t$-test, using $\nu=n-1$.
- Two independent samples with a common variance: use the pooled estimate 合并估计 and $\nu=n_1+n_2-2$:
Degrees of freedom, not $n$. One sample or paired: $\nu=n-1$. Two samples pooled: $\nu=n_1+n_2-2$. Reading the wrong row of the table is the most common error.
Two independent samples of sizes 8 and 11 are compared with a pooled t-test. How many degrees of freedom?
Сложность 2/5 · 140 монет + небольшой бонус за первое завершение сегодня
ν = n₁ + n₂ − 2 = 8 + 11 − 2 = 17.
Ten patients have their blood pressure measured before and after a drug. Which test is appropriate?
Сложность 1/5 · 100 монет + небольшой бонус за первое завершение сегодня
Each patient is measured twice, so the data are paired: test the differences with ν = 9.
Worked exam question: a paired t-test
Question (7 marks). Six students took a reading test before and after a course. The improvements $d$ (after − before) were $3,\ 5,\ -1,\ 4,\ 2,\ 5$. Test at the 5% level whether the course improves scores, stating an assumption.
- Assumption: the differences come from a normal distribution (B1).
- $H_0$: $\mu_d=0$; $H_1$: $\mu_d>0$ (B1).
- $\sum d=18$, so $\bar d=3$. $\sum d^2=80$, so $s^2=\dfrac{80-18^2/6}{5}=5.2$ (M1 A1).
- Test statistic (M1):
- $\nu=5$, one-tailed 5% critical value $2.015$. Since $3.22>2.015$, reject $H_0$: there is evidence at the 5% level that the course improves reading scores (A1).
Why answers lose marks: treating the data as two independent samples (they are paired, so use the differences); using $\nu=6$; or dividing by $n$ instead of $n-1$ for $s^2$.
Without looking back: list the steps of a paired t-test, from the assumption to the conclusion.
Follows the worked question.
You've got it
- small sample + unknown variance → use the $t$-distribution ($n-1$ degrees of freedom)
- CI for the mean: $\bar{x} \pm t\dfrac{s}{\sqrt{n}}$, with standard error $\dfrac{s}{\sqrt n}$
- paired: test the differences ($\nu=n-1$); two samples: pooled $s_p^2$, $\nu=n_1+n_2-2$
- state the normal assumption
How many degrees of freedom are used?
Сложность 1/5 · 100 монет + небольшой бонус за первое завершение сегодня
16 minus 1 is 15. Other options — 16 (the sample size); 9 (the lesson value for a sample of size 10).