Probability and Sampling Distributions

A sample statistic varies from sample to sample. Its sampling distribution describes that variation and provides the basis for standard errors, confidence intervals, and hypothesis tests.

Sample summaries

For observations $X_1,\ldots,X_n$, the sample mean is

\[\bar X=\frac{1}{n}\sum_{i=1}^n X_i.\]

The sample variance is

\[s^2=\frac{1}{n-1}\sum_{i=1}^n(X_i-\bar X)^2.\]

The divisor $n-1$ makes $s^2$ unbiased for the population variance under independent, identically distributed sampling.

Expectation and variance

For a discrete random variable,

\[E[X]=\sum_x xP(X=x),\]

and for a continuous random variable with density $f$,

\[E[X]=\int_{-\infty}^{\infty}xf(x)\,dx.\]

Variance can be written in either form:

\[\operatorname{Var}(X)=E[(X-E[X])^2]=E[X^2]-E[X]^2.\]

For constants $a$ and $b$,

\[E[aX+b]=aE[X]+b, \qquad \operatorname{Var}(aX+b)=a^2\operatorname{Var}(X).\]
Adding a constant changes location but not spread; multiplying by $a$ multiplies standard deviation by $ a $ and variance by $a^2$.

Standard error of the mean

If the observations are independent with common variance $\sigma^2$,

\[\operatorname{Var}(\bar X)=\frac{\sigma^2}{n}, \qquad SE(\bar X)=\frac{\sigma}{\sqrt n}\approx\frac{s}{\sqrt n}.\]

The $1/\sqrt n$ rate means that reducing the standard error by half requires four times as many observations.

Central limit theorem

Under regularity conditions, the standardized sample mean approaches a standard normal distribution:

\[\frac{\bar X-\mu}{\sigma/\sqrt n}\overset{d}{\longrightarrow}N(0,1).\]

For a sufficiently large sample this is used as the approximation

\[\frac{\bar X-\mu}{\sigma/\sqrt n}\approx N(0,1).\]

The CLT concerns the distribution of the sample mean, not necessarily the distribution of the raw observations.

Important distributions

If $X$ counts successes in $n$ independent Bernoulli trials,

\[X\sim\operatorname{Bin}(n,p), \qquad E[X]=np, \qquad \operatorname{Var}(X)=np(1-p).\]

Standardizing a normal random variable gives

\[Z=\frac{X-\mu}{\sigma}\sim N(0,1).\]

For a normal sample with unknown variance,

\[T=\frac{\bar X-\mu}{S/\sqrt n}\sim t_{n-1}.\]

The sample variance has the exact normal-sample relationship

\[\frac{(n-1)S^2}{\sigma^2}\sim\chi^2_{n-1}.\]

These last two statements are exact when the observations are normally distributed. Large-sample approximations can remain useful outside exact normality.

Connection to experiments

The same logic applies to treatment-control differences. The estimator $\bar Y_1-\bar Y_0$ has a distribution across repeated assignments, and Neyman Repeated-Sampling Inference uses its variance to form a standard error.

See