GSEM standardized coefficient and estimation

[Quynh Le]

Hi Statalist members,

I am using GSEM in Stata 13. I wonder if there is a way that I can calculate the standardized coefficient of my models (I have 3 models simultaneously)? Also how can I do a what-if simulation in STATA (change values of variables and see if they influence dependent variables)?

I am a beginner in structural equation modelling and followed the 4 commands to calculate good of fit test from Alan Acock book. It seems that they only work for SEM, not GSEM. Is there a way that I can find out how to improve my current model?

Thanks

[Marcos Almeida]

Hello Quynh,

Welcome to the Stata Forum / Statalist.

With regards to the standardized coefficients, you may get the answer from the Stata Manual, on page 450:

gsem cannot produce standardized coefficients.

With regards to differences between sem and gsem, we read in page 3:

[...] sem provides features not provided by gsem: easy testing for whether groups such as
males and females differ; estimation using observations with missing values under the assumption of
joint normality; goodness-of-fit statistics, modification indices, tests of indirect effects, and more; and
models fit using summary-statistic data.

Please prefer to write Stata, only the first letter in upper case, as recomended in FAQ.

With regards to a "better way to improve your model", I wish to kindly underline that the odds of getting helpful advice improve whenever data and command are shared.

Hopefully that helps.

[Quynh Le]

Hi Marcos,

Thanks a lot!

[Joaquin Carrascosa]

Hi, sory for reviving this post (not sure if it´s against the rules)

Could i just standarize the variables in gsem? (for linear models with quantitative variables only)

What if im using a factor variable in my model ? (in my case where this variable is involved i use a mlogit regression)

Or should i just work with unstandardized coefficientes ?

Thanks!