The Sampling Distribution of x-bar

A single person's height can be way above or below average. But average 40 people's heights, and the extremes tend to cancel out: one unusually tall person gets balanced by someone shorter. That averaging is why x-bar varies far less than individual values do. Take many different samples of size n from the same population, compute x-bar for each, and those means form their own distribution, the sampling distribution of x-bar. Its center is mu, the population mean, because samples don't systematically run high or low. Its spread is sigma/√n, called the standard error, where sigma is the population standard deviation. Notice the √n in the denominator: the standard deviation of x-bar is NOT sigma itself, because x-bar is an average of n values, and averaging cancels variation that a single observation can't. As n grows, that canceling gets more thorough, so the standard error shrinks, though never to zero.

This machinery sets up everything in the rest of the unit. Because sigma is almost never known in practice, it gets replaced by the sample standard deviation s, and dividing by s instead of sigma introduces extra uncertainty. That's precisely why the t-distribution, wider-tailed than the normal curve, takes over when building confidence intervals and running hypothesis tests for means. Understanding why x-bar's spread depends on both sigma and n explains why larger samples give tighter, more trustworthy estimates, and sets the stage for every t-procedure that follows.