Hi, I would like to add parallel analysis (based on normally distributed random data generation AND on permutations of the original raw data set - together with the scree-like parallel graph) to the wish list for the following versions. I guess that with the Monte Carlo facilities already present in Mplus it should be feasible. With ESEM, Mplus has already taken a great lead over other programs for EFA. With PA facilities, I guess it would make this lead even greater. Yes, various simple programs already exists to do PA on fortran, vista, SPSS, SAS but they dont always converge and all have inadequate missing data facilities (they mostly end up doing listwise deletion and, when you manage to avoid listiwise by imputing the data set, they often result in different results)... Anyhow, I guess Mplus could be first on this topic also !
Hi again Getting back to parallel analysis, there is two ways to doing it (step 1 is estimating the eigne values from the data, and step 2 is): (1) Estimating the eigen value from a simulated data set of random variables (same n as the real sample, same number of variables). How would you specify the Model population input of the Monte Carlo feature to obtain random variables ? This can surely be done in Mplus ? (2) The second way is by creating the "random variables" by permutations of the real data set. This probably cannot be done yet ?
I'll try that and let you know how the results compare to other methods of conducting PA (i.e. O'connor macros for SPSS). Thanks
EFried posted on Tuesday, September 18, 2012 - 9:24 am
Dear MPLUS team,
Will parallel analysis mentioned above by Alexandre Morin be implemented in MPLUS7? It would be very nice indeed to have yet another tool to combine it with the advantages MPLUS has over other software (e.g. missing data & ESEM) when it comes to factor extraction.