Hello. Please refer to input statement below. I am attempting to run a model in which I am regressing an outcome (bmi) on a latent factor (AccX) adjusting for covariates (dumdom dumca dumcub dumpr dumsa age sex educ2 income nhc), and I would additionally like to examine how the residual variance of one of the indicators of the latent factor (yrsus) relates to the outcome controlling for the latent factor and the covariates. My concern is that I may not be specifying this model correctly, as the association between the factor and outcome changes dramatically when comparing a model that does not include the relationship between the residual variance of the indicator and the outcome to a model that does include this relationship. This is occurring across multiple models examining various different outcomes. Do you know of any examples where this has been done and published previously and/or may you help me understand if I am specifying my model correctly? I am using the following input statement:
AccX BY yrsus usborn sashlanx sashsoc;
bmi ON AccX;
bmi ON dumdom dumca dumcub dumpr dumsa age sex educ2 income nhc;
I think you are referring to a relationship between the bmi outcome and a residual of a factor indicator, not a residual variance. To do this, see our FAQ "Regressing on a residual". FAQs are shown in the left margin of our homepage.