Dear list,
i'm trying to make sense of the variance of the specific/unique factors from a
factanal model. To demonstrate, suppose
fdata <- matrix( rnorm(100*5, mean=1, sd=2), nrow=100, ncol=5)
factmod <- factanal( fdata, factors=2, method="mle")
phi <- factmod$uniqueness
Now I thought the following should validate the variance of the specific factor,
ctr <-matrix( rep(t(factmod$center), nrow(fdata)), ncol=ncol(fdata), byrow=T)
fdata.demean <- fdata - ctr
lambda <- loadings(factmod)
factors <- factmod$scores
LF <- t(lambda %*% t(factors) )
resid <- fdata - ctr - LF
'resid' in last line is just the specific factors in this realization. My
question is shouldn't 'phi' be just
diag(var(resid))
Apparently it is not. What did i do wrong here?
Thanks for your insight in advance.
Cheers.
Horace W. Tso
|