Student's t-distribution

When you estimate a population mean from a small sample, you face a double uncertainty: you do not know the true mean AND you do not know the true spread — you have to estimate the standard deviation from the same handful of data. Student's t-distribution is the bell-like distribution that honestly accounts for this extra wobble, giving slightly fatter tails than the normal so your conclusions are appropriately cautious.

Formally, a t with k degrees of freedom arises as T = Z / sqrt(V/k), where Z is standard normal and V is an independent chi-squared with k degrees of freedom — exactly the structure of a sample mean standardized by an ESTIMATED standard deviation. It is symmetric and centred at 0 like the standard normal but has heavier tails: with few degrees of freedom the tails are noticeably fat (extreme t-values are more likely), and as the degrees of freedom k grow, the chi-squared part stabilizes and the t-distribution converges to the standard normal. So 'small sample' means use t; 'large sample' means t and normal nearly coincide.

The t-distribution is the engine of the t-test and the t-based confidence interval for a mean — the everyday tools when the population standard deviation is unknown, which is almost always. Its name is a quirk of history: William Gosset published under the pen name 'Student' while working at the Guinness brewery. The honest caveat: the t-test's validity rests on the data being roughly normal (or the sample being large enough for the CLT); for very small, badly skewed samples even the t can mislead.