Next: groupcovar() Up: Multivariate Macros Help File Previous: glsresids()   Contents


goodfit(s,lhat,psihat), REAL symmetric matrix lhat REAL vector or
  matrix, psihat REAL vector, all with no MISSING values

Keywords: factor analysis
goodfit(r, L, psi), where r is a p by p symmetric correlation or
variance-covariance matrix, L is a REAL p by m matrix of purported
factor loadings, and psi is a length p REAL vector of factor analysis
uniquenesses computes three measures of lack of fit of r to the factor
analytic approximation rhat = L %*% L' + dmat(psi).

If dev = vector(r - rhat) is the length p^2 vector of deviations of r
from rhat, then goodfit() returns
       vector(max(abs(dev)), sum(dev^2), sum(abs(dev)))

When L and psi have been computed by some factor analysis estimation
method, and one or more of the elements of the result is large it may
indicate lack of fit of r to the factor analytic model.

These are intended to describe how bad the fit is, not to test it.

Gary Oehlert 2003-01-15