Not a specific problem with CFA, I guess, but because I experienced the next problem doing a CFA, I've posted my problem here.
The question is simple: what is the maximum number of variables Mplus can handle doing a CFA (for example)?
I have a data file with over 600 variables (longitudinal data) and would like to do a CFA on it. But because one respondent takes more than one rule in the .DAT-file, Mplus thinks there are more respondents than there are. And so, CFA is not possible.
Is there a solution for this problem, or do I simply have to do the CFA with less variables? And if, how many then?
Nevertheless, even with 500 variables the problem still exists. Part of the problem is that when using for example 500 variables, the data of each respondent use more than one rule in the .DAT-file. This is what I meant before with Mplus that thinks there are more respondents then really are.
Is there are command by which one can tell Mplus that each two rules of data are part of one respondent (other words: each respondent containts two rule of data).
Mplus does not restrict data reading to one record per observation. If you are having problems reading the data, please send your input, data, output, and license number to firstname.lastname@example.org and I can take a look at it.
Jan Ivanouw posted on Thursday, June 05, 2008 - 7:00 am
As I understand it, this is a case of long versus wide data format for longitudinal data? Solutions are described in the Mplus manual
Dr. Muthen, I am trying to create SEM with 105 variables. I investigated the latent variables separately to make sure I have good fit models before I develop the SEM. However, when I try to test the measurement model with the latent variables, I receive a message that I should use NSERROR NCHISQUARE to reduce computation time. But this will not allow for model fit incidences to be computed. Do you suggest another method to test the measurement models/and SEM model?