Make sense of data, one idea at a time.

First identify what you want to learn: compare groups, estimate an association, or test a proportion. Then classify the outcome as numeric, categorical, or ordinal, and note whether observations are independent or paired.

For example, comparing a numeric outcome between two independent groups often points to a two-sample t-test, while paired measurements call for a paired method. For categorical counts, a chi-square test or Fisher's exact test may fit. Check assumptions and sample size before deciding.

A p-value is the probability, assuming the null hypothesis and model assumptions are true, of observing a result at least as incompatible with the null as the one you saw. It is not the probability that the null hypothesis is true.

Read it alongside the effect estimate, confidence interval, study design, and practical importance. A small p-value can flag a mismatch with the null model, but it does not measure the size or value of an effect.

The mean uses every value and is useful when the distribution is reasonably balanced and the arithmetic average is meaningful. A few extreme observations can pull it away from where most values sit.

The median is the midpoint after sorting and is more resistant to extreme values. For skewed data such as income, report the median with a measure of spread like the interquartile range. A histogram or box plot helps show why one summary may be more informative.