

Analysis with binary predictor 

Message/Author 

Maša Vidmar posted on Monday, December 26, 2011  12:24 pm



I have SEM model with four dependent latent variables (at two time points) and gender as predictor. I am not sure how to run the analysis and interpret the results. My syntax: beh1 by x1 x2 x3 x4 beh2 by x1_2 x2_2 x3_2 x4_2 ach1 by z1 z2 z3 ach2 by z1_2 z2_2 z3_2 ach2 on ach1 beh1; beh2 on beh1 ach1; beh1 on gender; ach1 on gender; My questions: 1. Is it ok to use ml as estimator? Should I be using some other method? 2. Should I be looking at the STDY and not STDYX part of the output(it makes no sense to standardize gender, right?)? 3. How to interpret coefficients? In my case, path from gender to behavior is n.s., path from gender to achievement is sig. and negative. Is interpretation the same as in regression; when gender is changed for one unit (ie. from 1female to 2male) achievement is lower for .15 points? Should this value .15 point be taken from the nonstandardized or standardized part of the output? Another questions, ona a different note. In a similar model to the one above, I would like to use predictor that is not categorical, but can take only three (ordinal) values (years is preschool: 0, 3 or 5). Should I treat it as categorical and create dummy variables? Or do I enter it in the model as if it was interval? I am worried about the interpretation in both cases. 


1. Yes. 2. Use StdY. See the STANDARDIZED option in the user's guide for further information. 3. The interpretation is the same as for regular regression. An ordered categorical predictor can be treated as continuous or turned into a set of dummy variables. 

Jan Ivanouw posted on Monday, November 26, 2012  11:33 am



I have the problem that the result file in version 7 only gives StdXY and Std, but not StdY. What should I use for the binary predictors? 


Create StdY by dividing StdYX by the standard deviation of x. 

Back to top 

