Skip to content

Factor analysis and cfa with asymptotically distributed data

2 messages · "Jeanine Grütter", Mark Difford

Original
"Jeanine Grütter"
# 
I have friendship data which is strong skewed. So it doesn't make sense to 
use maximum likelihood methods for fa and cfa.
But I couldn't find any function for asymptotically distributed data for doing a factor analysis. Only: apca() but there is no possibility to allow for factor correlations. 

The same problem is with sem() I couldn't get any solutions for my model because of the distribution.
I get the error: 
In sem.default(ram = ram, S = S, N = N, param.names = pars, var.names = vars,  :
  Could not compute QR decomposition of Hessian.
Optimization probably did not converge.

Does someone know how to do fa and cfa with strong skewed data?

Thanks a lot for helping!

Jane
--
Mark Difford
# 
Jane,
Your best option might be to use a robustly estimated covariance matrix as
input (see packages robust/robustbase).

Or you could turn to packages FAiR or lavaan (maybe also OpenMx). Or you
could try soft modelling via package plspm.

Regards, Mark.