What is the difference in Std and StdXY in MPlus output for a 2-level CFA-model with ordinal observed variables? On the within-level Std>1 and StdXY<1, on the between-level Std looks quite reasonable and StdXY is fixed to 1.
Which of the standardized parameters should I report when describing the results of the 2-level-model, Std or StdXY?
bmuthen posted on Sunday, December 18, 2005 - 6:01 pm
The User's Guide discusses how to interpret these columns. The Std column standardizes with respect to the latent variables only so here it is quite common that the values are > 1 because the non-latent variables (in your case the factor loadings) may have variance much greater than 1. For factor loadings of ordinal indicators, the StdYX column standardizes with respect to both the factors and the latent response variables and then it is more common to find values < 1. Note, however, that this is not always the case with correlated factors (see for example notes by Joreskog on the SSI web site).
If you report loadings and want to report standardized values in addition to the raw values, I would choose StdYX.
I'm doing a twolevel analysis (random intercept only) with pupils clustered in classes. Most of my predictors on level 1 are manifest and continous variables (e.g. IQ, Motivation, Academic Self Concept) but I have "sex" as a dichotomous variable. I want to report the regression coefficients in my study. Is ist right to take the regression coeffecients for the continous variables form the stdyx output and the regression coefficient for the dichotomous variable from the stdy output?