Lessons and exercises follow the school system and stay in its language.Les leçons et les exercices suivent le système scolaire et restent dans sa langue.Lektionen und Übungen folgen dem Schulsystem und bleiben in seiner Sprache.

Sampling distributions

Print the topic

Every lesson of the topic, from the first year to the last, ready for paper. Exercise sets are lettered A, B, C… and answered at the end.

To save a PDF, choose “Save as PDF” in the print dialog.

1. Sampling variability and sampling distributions

Warm-up

Three quick questions from earlier lessons.

  1. A researcher measures the sleep of 376 randomly chosen teenagers and finds a mean of 7.1 hours. What is 7.1?
    1. A statistic: it describes the sample
    2. A parameter: it describes all teenagers
    3. The population
  2. To learn how much US college students spend on books, a team surveys 671 students at 20 colleges. What is the population?
    1. All US college students
    2. The 671 students surveyed
    3. The 20 colleges
  3. Census records show that 52% of all households in a town own a car. What is 52%?
    1. A parameter: it describes the whole population
    2. A statistic
    3. A sample proportion

Discover

Five classmates each survey 20 random students about a new rule. Will they all get the same percentage in favor?

  1. No, the percentages will vary
  2. Yes, if they all sample at random

Key idea

A statistic has a distribution of its own.

  1. Step 110 samples: a few values of p̂, scattered.
  2. Step 2100 samples: a shape appears.
  3. Step 31,000 samples (scaled): a mound centered near p — the sampling distribution.

The sampling distribution of a statistic is its distribution over all possible random samples of the same size. Its center tells us about bias; its spread (standard error) shrinks as the sample size grows.

Worked example

Imagine 1,000 samples of 50 students and the p̂ of each.

A histogram of p̂ values
Plot the 1,000 values.
Centered near the true p
Where it is centered.
Narrower: less variability
With samples of 200 instead.

Your turn, guided

Your turn: by what factor does SD(x̄) change when nn goes from 25 to 100?

It is halved: factor 12\tfrac{1}{2}
√25/√100.

Practice ★

  1. Many random samples of size 161 are taken from the same population. Which quantity changes from sample to sample?
    1. The sample proportion p̂ (a statistic)
    2. The population proportion p (a parameter)
    3. The sample size
  1. Compared with samples of size 73, samples of size 292 give sample means that…
    1. vary less around the population mean
    2. vary more
    3. are more biased

Watch out!

Practice ★★

  1. What is the sampling distribution of the sample mean for samples of size 25?
    1. The distribution of x̄ over all possible samples of size 25
    2. The distribution of the 25 values in one sample
    3. The distribution of the population

Exit ticket

Three questions to check you have got it.

  1. Many random samples of size 161 are taken from the same population. Which quantity changes from sample to sample?
    1. The sample proportion p̂ (a statistic)
    2. The population proportion p (a parameter)
    3. The sample size
  2. Many random samples of size 224 are taken from the same population. Which quantity changes from sample to sample?
    1. The sample proportion p̂ (a statistic)
    2. The population proportion p (a parameter)
    3. The sample size
  3. Many random samples of size 490 are taken from the same population. Which quantity changes from sample to sample?
    1. The sample proportion p̂ (a statistic)
    2. The population proportion p (a parameter)
    3. The sample size

Summary card

Sampling variability

  • Statistics vary from sample to sample.
  • Sampling distribution: a statistic over all samples.
  • Larger nn: less variability.

Watch out: data distribution vs sampling distribution.

One more step is optional: a challenge. Try it, or finish the lesson now.

Challenge ★★★

  1. Samples of size 22 give sample means with a spread (standard deviation) of ss. Samples of size 352 are taken instead. By what factor does the spread shrink?

2. The sampling distribution of a sample proportion

Warm-up

Three quick questions from earlier lessons.

  1. In a random sample of 100, p̂ = 3/10. Build a 95% interval with z* = 1.96.
    1. The standard error:   ?\;?
    2. The margin of error:   ?\;?
    3. The lower end of the interval:   ?\;?
    4. The point estimate (center of the interval):
    5. With a sample 4 times as large, the standard error is divided by:
  2. A machine part is faulty with probability 1/5, independently. Four parts are tested; X counts the faulty ones.
    1. P(X=0)P(X = 0):   ?\;?
    2. P(X≥1)P(X \ge 1):   ?\;?
    3. E(X)E(X):   ?\;?
    4. σ(X)\sigma(X):
    5. The probability that the first two parts are faulty and the other two are not:
  3. X is binomial with n = 30 and p = 1/10. Find E(X).

Discover

In a city 40% of adults cycle. In random samples of 100, around which value will p̂ be centered?

  1. 0.40
  2. 40

Key idea

Mean, spread and shape of p̂.

For random samples of size nn from a population with proportion pp: μp^=p\mu_{\hat{p}} = p and σp^=p(1−p)n\sigma_{\hat{p}} = \sqrt{\dfrac{p(1 - p)}{n}} (if nn is at most 10% of the population). The shape is approximately normal when np≥10np \ge 10 and n(1−p)≥10n(1 - p) \ge 10.

Worked example

p=0.4p = 0.4, n=100n = 100.

μp^=0.4\mu_{\hat{p}} = 0.4
Center.
σp^=0.0024≈0.049\sigma_{\hat{p}} = \sqrt{0.0024} \approx 0.049
√(0.4 × 0.6/100).
Approximately normal
40 ≥ 10 and 60 ≥ 10.

Your turn, guided

Your turn: p=0.5p = 0.5, n=100n = 100. Find σp^\sigma_{\hat{p}}.

0.050.05
√(0.25/100).

Practice ★

  1. The population proportion is p = 8/10. Find the standard deviation of p̂ for random samples of size 83.
  2. The population proportion is p = 2/10. Find the standard deviation of p̂ for random samples of size 227.
  1. The population proportion is p = 5/10. What is the mean of the sampling distribution of p̂ for samples of size 351?

Watch out!

Practice ★★

  1. The population proportion is p = 1/8. What is the smallest sample size n for which np ≥ 10 and n(1 − p) ≥ 10?
  2. The population proportion is p = 1/2. What is the smallest sample size n for which np ≥ 10 and n(1 − p) ≥ 10?

Exit ticket

Three questions to check you have got it.

  1. The population proportion is p = 8/10. Find the standard deviation of p̂ for random samples of size 83.
  2. The population proportion is p = 2/10. Find the standard deviation of p̂ for random samples of size 227.
  3. The population proportion is p = 4/10. Find the standard deviation of p̂ for random samples of size 394.

Summary card

Sampling distribution of p̂

μ=p\mu = p, σ=p(1−p)/n\sigma = \sqrt{p(1 - p)/n}; normal if np,n(1−p)≥10np, n(1 - p) \ge 10.

Watch out: proportion, not count.

One more step is optional: a challenge. Try it, or finish the lesson now.

Challenge ★★★

  1. For p = 0.5, how large must the sample be for the standard deviation of p̂ to be 1100\dfrac{1}{ 100 }?

3. The sampling distribution of a sample mean

Warm-up

Three quick questions from earlier lessons.

  1. In a random sample of 100, p̂ = 3/10. Build a 95% interval with z* = 1.96.
    1. The standard error:   ?\;?
    2. The margin of error:   ?\;?
    3. The lower end of the interval:   ?\;?
    4. The point estimate (center of the interval):
    5. With a sample 4 times as large, the standard error is divided by:
  2. X and Y are independent with SD(X) = 3 and SD(Y) = 4. Find SD(X + Y).
  3. X and Y are independent with SD(X) = 12 and SD(Y) = 5. Find SD(X − Y).

Discover

Heights have SD 10 cm. Is the SD of the mean of 25 random heights also 10 cm?

  1. No, it is 2 cm
  2. Yes

Key idea

Mean and spread of x̄.

For random samples of size nn from a population with mean μ\mu and SD σ\sigma: μxˉ=μ\mu_{\bar{x}} = \mu and σxˉ=σn\sigma_{\bar{x}} = \dfrac{\sigma}{\sqrt{n}}. If the population is normal, xˉ\bar{x} is exactly normal.

Worked example

μ=170\mu = 170, σ=10\sigma = 10, n=25n = 25. How unusual is xˉ=174\bar{x} = 174?

σxˉ=2\sigma_{\bar{x}} = 2
10/5.
z=2z = 2
(174 − 170)/2.
Unusual (about 2.5% are higher)
Beyond 2 SDs.

Your turn, guided

Your turn: σ=12\sigma = 12, n=16n = 16. Find σxˉ\sigma_{\bar{x}}.

33
12/4.

Practice ★

  1. A population has standard deviation σ = 26. Find the SD of the sample mean for samples of size 9.
  2. A population has standard deviation σ = 84. Find the SD of the sample mean for samples of size 400.
  3. A population has standard deviation σ = 29. Find the SD of the sample mean for samples of size 4.

Watch out!

Practice ★★

  1. Samples of size 26 give a certain SD of x̄. By what factor must the sample size be multiplied to halve it?
  1. A population has mean 76 and SD 5. A random sample of 4 has mean 77. Find the z-score of this sample mean.
  2. A population has mean 62 and SD 14. A random sample of 16 has mean 69. Find the z-score of this sample mean.

Exit ticket

Three questions to check you have got it.

  1. A population has standard deviation σ = 26. Find the SD of the sample mean for samples of size 9.
  2. A population has standard deviation σ = 84. Find the SD of the sample mean for samples of size 400.
  3. A population has standard deviation σ = 29. Find the SD of the sample mean for samples of size 4.

Summary card

Sampling distribution of x̄

μxˉ=μ\mu_{\bar{x}} = \mu, σxˉ=σ/n\sigma_{\bar{x}} = \sigma/\sqrt{n}.

Watch out: divide by n\sqrt{n}, not nn.

One more step is optional: a challenge. Try it, or finish the lesson now.

Challenge ★★★

  1. For which sample size is the standard deviation of x̄ equal to σ2\dfrac{\sigma}{ 2 }?

4. The central limit theorem

Warm-up

Three quick questions from earlier lessons.

  1. A population has standard deviation σ = 26. Find the SD of the sample mean for samples of size 9.
  2. Samples of size 26 give a certain SD of x̄. By what factor must the sample size be multiplied to halve it?
  3. A population has mean 76 and SD 5. A random sample of 4 has mean 77. Find the z-score of this sample mean.

Discover

Waiting times at a clinic are strongly right-skewed. You average the times of 50 random patients, many times over. What shape do these averages have?

  1. Approximately normal
  2. Right-skewed

Key idea

Averages tend to be normal.

Central limit theorem: for random samples of size nn from any population with finite SD, the sampling distribution of xˉ\bar{x} is approximately normal when nn is large (a common guide: n≥30n \ge 30, more for strong skew). It is about the distribution of sample means, not of the data.

Worked example

Three situations.

Any nn: xˉ\bar{x} is normal
Normal population.
xˉ\bar{x} approximately normal (CLT)
Skewed population, n = 60.
xˉ\bar{x} still skewed: CLT does not apply yet
Skewed population, n = 4.

Your turn, guided

Your turn: a population with σ=15\sigma = 15 and n=36n = 36. The SD of the (approximately normal) xˉ\bar{x}.

2.52.5
15/6.

Practice ★

  1. Household incomes are strongly right-skewed. What is the shape of the sampling distribution of x̄ for random samples of 52 households?
    1. Approximately normal, by the central limit theorem
    2. Right-skewed like the population
    3. Uniform
  1. A population is strongly skewed. Is the sampling distribution of x̄ for samples of size 3 approximately normal?
    1. Not necessarily: n = 3 is too small for the CLT to apply to a strongly skewed population
    2. Yes, sample means are always normal
    3. No, sample means are never normal

Watch out!

Practice ★★

  1. Which statement is the central limit theorem, for random samples of size 166?
    1. The distribution of sample means is approximately normal
    2. The data in a large sample are approximately normal
    3. The population becomes normal when n is large

Exit ticket

Three questions to check you have got it.

  1. Household incomes are strongly right-skewed. What is the shape of the sampling distribution of x̄ for random samples of 52 households?
    1. Approximately normal, by the central limit theorem
    2. Right-skewed like the population
    3. Uniform
  2. Household incomes are strongly right-skewed. What is the shape of the sampling distribution of x̄ for random samples of 73 households?
    1. Approximately normal, by the central limit theorem
    2. Right-skewed like the population
    3. Uniform
  3. Household incomes are strongly right-skewed. What is the shape of the sampling distribution of x̄ for random samples of 72 households?
    1. Approximately normal, by the central limit theorem
    2. Right-skewed like the population
    3. Uniform

Summary card

Central limit theorem

Large nn: xˉ\bar{x} approximately normal, whatever the population.

Watch out: it is about means, not about the data.

One more step is optional: a challenge. Try it, or finish the lesson now.

Challenge ★★★

  1. A skewed population has mean 72 and SD 6. By the central limit theorem, x̄ for samples of 36 is approximately normal. What is its standard deviation?

5. Sampling distributions of differences

Warm-up

Three quick questions from earlier lessons.

  1. The population proportion is p = 8/10. Find the standard deviation of p̂ for random samples of size 83.
  2. The population proportion is p = 5/10. What is the mean of the sampling distribution of p̂ for samples of size 351?
  3. The population proportion is p = 1/8. What is the smallest sample size n for which np ≥ 10 and n(1 − p) ≥ 10?

Discover

Two independent sample means have SDs 3 and 4. What is the SD of their difference?

  1. 5
  2. 1

Key idea

Differences between independent samples.

For independent samples: μp^1−p^2=p1−p2\mu_{\hat{p}_1 - \hat{p}_2} = p_1 - p_2, σp^1−p^2=p1(1−p1)n1+p2(1−p2)n2\sigma_{\hat{p}_1 - \hat{p}_2} = \sqrt{\dfrac{p_1(1 - p_1)}{n_1} + \dfrac{p_2(1 - p_2)}{n_2}}; and σxˉ1−xˉ2=σ12n1+σ22n2\sigma_{\bar{x}_1 - \bar{x}_2} = \sqrt{\dfrac{\sigma_1^2}{n_1} + \dfrac{\sigma_2^2}{n_2}}.

Worked example

p1=0.5p_1 = 0.5, p2=0.3p_2 = 0.3, n1=n2=100n_1 = n_2 = 100.

μ=0.2\mu = 0.2
Center.
σ=0.0046≈0.068\sigma = \sqrt{0.0046} \approx 0.068
0.0025 + 0.0021.

Your turn, guided

Your turn: σ12/n1=9\sigma_1^2/n_1 = 9 and σ22/n2=16\sigma_2^2/n_2 = 16. Find σxˉ1−xˉ2\sigma_{\bar{x}_1 - \bar{x}_2}.

55
√25.

Practice ★

  1. Independent samples of size 4 come from populations with SDs 36 and 48. Find the SD of x̄₁ − x̄₂.
  2. Independent samples of size 4 come from populations with SDs 18 and 24. Find the SD of x̄₁ − x̄₂.
  1. Independent random samples of 100 are taken from two populations with proportions 8/10 and 1/10. Find the SD of p̂₁ − p̂₂.

Watch out!

Practice ★★

  1. When are the variances of x̄₁ and x̄₂ added to get the variance of x̄₁ − x̄₂ (samples of 66)?
    1. When the two samples are independent
    2. Only when the sample sizes are equal
    3. Always, even for paired measurements

Exit ticket

Three questions to check you have got it.

  1. Independent samples of size 4 come from populations with SDs 36 and 48. Find the SD of x̄₁ − x̄₂.
  2. Independent samples of size 4 come from populations with SDs 18 and 24. Find the SD of x̄₁ − x̄₂.
  3. Independent samples of size 4 come from populations with SDs 54 and 72. Find the SD of x̄₁ − x̄₂.

Summary card

Differences

Independent samples: means subtract, variances add.

Watch out: paired data are not independent.

One more step is optional: a challenge. Try it, or finish the lesson now.

Challenge ★★★

  1. Independent samples of size 9 come from populations with SDs 15 and 20. Find the SD of x̄₁ − x̄₂.

6. Bias, variability and the 10% condition

Warm-up

Three quick questions from earlier lessons.

  1. The population proportion is p = 8/10. Find the standard deviation of p̂ for random samples of size 83.
  2. The population proportion is p = 5/10. What is the mean of the sampling distribution of p̂ for samples of size 351?
  3. The population proportion is p = 1/8. What is the smallest sample size n for which np ≥ 10 and n(1 − p) ≥ 10?

Discover

A scale always reads 2 kg too heavy, but very consistently. Is it a good scale?

  1. No, it is biased
  2. Yes, it is very consistent

Key idea

Good estimators are unbiased and precise.

An estimator is unbiased if the mean of its sampling distribution equals the parameter (p^\hat{p} and xˉ\bar{x} are unbiased). Its variability shrinks as nn grows. The SD formulas assume independence; when sampling without replacement, keep nn at most 10% of the population.

Worked example

A town of 6,000 households.

Largest sample: 600 households
10% of 6,000.
A larger random sample
To shrink variability.
A larger sample does not fix a biased method
Bias comes from the method.

Your turn, guided

Your turn: population 2,500. The largest sample meeting the 10% condition.

250250
10% of 2,500.

Practice ★

  1. What does it mean that p̂ is an unbiased estimator of p (samples of 40)?
    1. The mean of its sampling distribution equals p
    2. Every sample gives p̂ = p
    3. Its variability is small
  1. A school has 5000 students. Sampling without replacement, what is the largest sample size that meets the 10% condition?
  2. A school has 28900 students. Sampling without replacement, what is the largest sample size that meets the 10% condition?

Watch out!

Practice ★★

  1. A survey method gives estimates that cluster tightly, but far from the true value. How would you describe it (samples of 116)?
    1. Biased, with low variability
    2. Unbiased, with low variability
    3. Unbiased, with high variability

Exit ticket

Three questions to check you have got it.

  1. What does it mean that p̂ is an unbiased estimator of p (samples of 40)?
    1. The mean of its sampling distribution equals p
    2. Every sample gives p̂ = p
    3. Its variability is small
  2. What does it mean that p̂ is an unbiased estimator of p (samples of 187)?
    1. The mean of its sampling distribution equals p
    2. Every sample gives p̂ = p
    3. Its variability is small
  3. What does it mean that p̂ is an unbiased estimator of p (samples of 106)?
    1. The mean of its sampling distribution equals p
    2. Every sample gives p̂ = p
    3. Its variability is small

Summary card

Bias and variability

  • Unbiased: centered on the parameter.
  • Variability: shrinks with nn.
  • 10% condition for independence.

Watch out: large does not mean unbiased.

One more step is optional: a challenge. Try it, or finish the lesson now.

Challenge ★★★

  1. Samples of 43 are drawn without replacement. What is the smallest population size that meets the 10% condition?

7. The square-root law

Discover

A sample of 100 gives a standard error of 4 for the mean. You want a standard error of 2. Is a sample of 200 enough?

  1. No: you need 400, four times as many
  2. Yes: double the sample, half the error

Key idea

The square-root law.

The standard error of a sample mean is

SE=σnSE = \dfrac{\sigma}{\sqrt{n}}.

To divide the SE by kk, multiply the sample size by k2k^2.

Worked example

σ=15\sigma = 15. Find the SE for n=25n = 25 and the nn that gives SE=1SE = 1.

SE=155=3SE = \dfrac{15}{5} = 3
n=25n = 25.
n=225n = 225
SE divided by 3: n×9n \times 9.

Your turn, guided

Your turn: σ=12\sigma = 12, n=16n = 16. Find the SE.

33
1216\dfrac{12}{\sqrt{16}}.

Practice ★

  1. A population has σ=27\sigma = 27. Find the standard error of the mean for samples of size 25.
  2. A population has σ=17\sigma = 17. Find the standard error of the mean for samples of size 100.
  3. A population has σ=3\sigma = 3. Find the standard error of the mean for samples of size 9.

Practice ★★

  1. A sample of 182 gives some standard error. What sample size divides that standard error by 5?
  2. A sample of 122 gives some standard error. What sample size divides that standard error by 4?

Exit ticket

Two questions to check.

  1. A population has σ=27\sigma = 27. Find the standard error of the mean for samples of size 25.
  2. A population has σ=17\sigma = 17. Find the standard error of the mean for samples of size 100.

Summary card

Square-root law

  • SE=σnSE = \dfrac{\sigma}{\sqrt{n}}.
  • Half the error costs four times the data.

This is why polls of about 1,000 people are common: going much further costs a lot for little gain.

8. History of sampling distributions: de Moivre, Laplace and Gosset

Discover

Toss a fair coin 1,000 times. The number of heads varies from one try to the next. In 1733, Abraham de Moivre found the shape of that variation. Which shape?

  1. A bell curve: the normal distribution
  2. A flat line: every count is equally likely

Key idea

Why sample statistics are predictable.

17501800185019001733de Moivre1733de Moivre
  1. Step 117501800185019001733de Moivre1733de MoivreAbraham de Moivre 1667–1754Joseph Highmore, 1736. Public domain13 November 1733: de Moivre approximates the binomial by the normal curve.
  2. Step 217501800185019001733de Moivre1812Laplace1733de Moivre1812LaplacePierre-Simon Laplace 1749–1827Jean-Baptiste Paulin Guérin, 1838. Public domain1812: Laplace's Analytic Theory of Probability.
  3. Step 317501800185019001733de Moivre1812Laplace1906Gosset1733de Moivre1812Laplace1906GossetWilliam Sealy Gosset 1876–1937Unknown artist. Public domain1906–07: Gosset studies the sampling distribution of the mean.

On 13 November 1733 de Moivre published, in a Latin pamphlet, the approximation of the binomial distribution by the normal distribution for a large number of trials. Laplace's Analytic Theory of Probability (1812) applied probability to errors in observations. In 1906–07 William Gosset studied the sampling distribution of the mean and standard deviation in Pearson's laboratory.

Worked example

De Moivre's idea with numbers: 400 tosses of a fair coin.

μ=200\mu = 200, σ=400×0.5×0.5=10\sigma = \sqrt{400 \times 0.5 \times 0.5} = 10
Mean and SD of the count.
between 180 and 220 heads
About 95 % of tries fall within 2 SD.

Practice ★

  1. What did de Moivre publish in 1733?
    1. The normal approximation to the binomial distribution
    2. The t-test for small samples
    3. Confidence intervals
  1. A fair coin is tossed 16 times. Find the standard deviation of the number of heads.
  2. A fair coin is tossed 100 times. Find the standard deviation of the number of heads.
  1. Same coin, 100 tosses. About 95 % of the counts lie within 2 SD of the mean. Give the upper limit.

Summary card

Sampling distributions today

  • Factories check samples of products, not every item.
  • Polls predict elections from about a thousand people.
  • Medical trials compare average results of groups.

Sources: MacTutor biographies of de Moivre, Laplace and Gosset (see "Sources and credits").