Statistical tests for proportions in matched groups

From Ganfyd

(This is a sub-page of the Medical statistics page.)

McNemar test

Used for proportions in matched groups.

Situation in which McNemar test can be used

Two individually matched groups, each of size N (N pairs of observations); qualitative binary variable with unknown population proportions π₁ and π₂ in one category of the variable. This category is denoted by a + sign below. Often the groups are cases and controls in a pair-matched case-control study, and the binary variable is exposure/non-exposure to a risk factor.

Assumptions/requirements for McNemar test:

None

Null hypothesis for McNemar test:

That the proportions π₁ and π₂ are equal.

Method for McNemar test:

Lay out the data in a two-by-two table as

where the classification is based on N pairs of observations, and the + and - signs refer to the categories of binary variable.