How to calculate the number of parameters involved in a SEM model

[Federico Tedeschi]

Dear Statalisters,
I’m using Stata 15.1.
I have used the “sem” command to estimate a mediation-analysis model where both the mediator and the outcome are latent factors. The problem is that my observed variables are so many that I get the error message “matsize too small”. Unfortunately, I have Stata IC so even increasing it to the maximum "matsize" allowed (800) I can’t solve this problem. I’d like to know how I can calculate the number of parameters that my model needs to estimate (so I can see how I can simplify it in order to build a one with less than 800 parameters). Also, I ask you whether you know either alternative commands or turn-arounds in Stata to deal with situations like that (mediation analysis with latent factors, estimated using a huge number of observed variables).

[Richard Williams]

R is free. If you are desperate enough to learn it, you could try the add-on lavaan routine.

As for Stata, you could try adding variables gradually and see how far you can take it.

You may be a lot happier, though, if you can find a way to upgrade to Stata/SE.

[Federico Tedeschi]

Thank you, Richard. I haven't been using R with real data for 14 years, but maybe this is a good opportunity to re-learn it.

[Phil Bromiley]

If you value your time, paying for SE may make a lot more sense than moving to R.