I am running an SEM using ML estimator. One of the paths in the model is not significant in the MODEL RESULTS. It is close to the borderline (p=0.058). However, this path is significant in the STANDARDIZED MODEL RESULTS (STDYX Standardization). The p value is 0.026. Can I report the standardized result? Is there a general rule on which one should be reported? Thanks a lot for any guidance.
You should report either raw or standardized results not a combination of both. The difference in significance is due to raw and standardized parameters having different sampling distributions. When you are doing so many tests at the same time, you should be conservative in your interpretation.
gibbon lab posted on Sunday, June 03, 2012 - 6:41 pm
Hi Professor Muthen,
Is there a paper or book where I can find the exact formula/equations that are used to generate the standardized results? Thanks a lot.
See Technical Appendix 3 on the website and the STANDARDIZED option in the user's guide.
Kirill Fayn posted on Thursday, August 22, 2019 - 7:23 am
Hello Professor Muthen,
I am running a latent change model (change between two time points) and predicting change with a continuous variable. The variable is a rating from 0-100 and has a mean of 22 and a SD of 23.16. For ease of interpretation I used the STANDARDIZE function (see code below). For some reason this changes my results drastically. Any idea why that might be?
Here is the code, thanks in advance for your help.
DEFINE: X1 = prep1; X2 = prep2;
standardize Shame_post ;
MODEL: LCS21X by X2@1; X2 @0; LCS21X ON X1; [X2 @0] ; X2 on X1@1;
LCS21X on Shame_post bfi_co_mean; OUTPUT: TECH1 TECH4 stand ;
I don't know what you mean by the results changing. The standardized estimates are different of course than the unstandardized. I think you need to send the two outputs to Support along with your license number.