Assessing independence, based on a sample

Independence is an important special case of models for bivariate data. However it is a property of the joint population probabilities and in most practical situations these are unknown.

We must assess independence from a sample of individuals — a contingency table.

Skiing injuries

Estimated cell counts under independence

If independence is assumed, we can find a pattern of cell counts that best corresponds to independence, based only on the margins of the table.

Skiing injuries