AP Statistics Section 10.2: Estimating a Population Mean
Example: Sitting vs. Standing Pulse Rates
Procedures that are not affected by lack of normality of the population are said to be robust. Translated: if we find that the condition regarding normality is not met, that our confidence interval will still be fairly accurate because the t-procedures are fairly robust. This does not guard against outliers as the confidence intervals are non-resistant. The CLT ensures the robustness. When n is small, it is more important that the data is an SRS than normally distributed.