Thursday, March 29, 2018

sample size too small to run CFA in SEM

CFA model, 16-item and 3 factor.
across 6 groups.
One group has only 21 cases.

the ratio of 21 cases for 16 items is far too small to estimate a stable covariance matrix to put into any size set of simultaneous equations. There's 120 pairs of similarity estimates to estimate, for
which 21 does not provide enough variability to produce stochastic estimation, plus the 16 individual variances.

No comments: