Bartlett's Chi-square for Canonical Correlation

From Q
Jump to: navigation, search

Tests for independence between two sets of Number - Multi questions, containing p and q variables respectively. The test statistic is:

 V = -((n - 1.5) - (p + q) / 2.0)\sum^m_{i=1} \log(1 - r^2_i)

where

n is the Effective Sample Size,
 p \approx \Pr(\chi^2_{pq} \ge \Lambda).

See also

Multivariate Tests