Cochran の Q 検定

Use Cochran's Q test when analyzing repeated measures with binary outcomes across three or more related groups. It extends the McNemar test, which handles only two related samples, making Cochran's Q more versatile for scenarios where the same subjects are measured under multiple conditions or time points. This test is ideal for medical research, behavioral studies, and agricultural applications involving yes/no or presence/absence data.

Cochran's Q test requires binary responses (0 or 1, yes or no), random sampling from the population, repeated measures where the same subjects are measured under each condition, and independence of individuals within each condition. Although outcomes may be dependent across conditions because the same subjects are used, violating these assumptions can lead to inaccurate conclusions, so verify data meets these criteria before applying the test.

The test statistic Q is calculated using an equation that incorporates the number of conditions (a), the number of subjects (b), and row and column totals from the binary data. In the plant example, Q was computed with a = 4 months and b = 24 saplings, where each sapling received a score of 1 for intact floral buds or 0 for absent buds. The resulting Q value is then compared against a chi-square distribution.

Cochran's Q test statistic approximates the chi-square distribution with degrees of freedom equal to the number of conditions minus one (a-1). Researchers compare the calculated Q value against the chi-square critical value at a chosen significance level, such as 0.05. If Q exceeds the critical value, the null hypothesis is rejected, indicating significant differences in binary outcomes across the related groups.

Cochran's Q test is widely used in medical research to evaluate treatment effectiveness on patient outcomes, behavioral studies to assess changes under different conditions, agricultural research to test treatment impacts on plant disease presence, and psychological experiments to investigate intervention effects on symptoms. Its flexibility in handling multiple related samples with binary outcomes makes it valuable for analyzing repeated measures data across diverse scientific fields.

Cochran's Q test specifically handles repeated measures by allowing the same subjects to be measured under multiple conditions while maintaining binary outcome data. Unlike tests that assume independence across all observations, Cochran's Q acknowledges that outcomes are dependent across conditions because identical subjects are used. This design makes it particularly powerful for tracking whether a binary attribute changes across different treatments or time points in the same individuals.

In the plant study example, 24 saplings were observed over four months to determine if floral bud presence changed when spiders were removed. Each sapling received a binary score monthly: 1 for intact buds, 0 for absent buds. Cochran's Q test analyzed whether the proportion of saplings with floral buds differed significantly across the four months, revealing the effect of pest damage on plant reproduction across the observation period.