

Latent Factor Means for Probit Plotting 

Message/Author 


I need the means for endogenous latent factors so I can plot predicted probabilities for a probit regression using WLSMV. In the past, you've said latent factors have means set to 0, except since I have covariates in the model, the estimated means are different than 0. Should I be plotting the estimated means, or is 0 still accurate despite the output in Tech4 and inclusion of covariates? 


If you have covariates influencing the factors, the factor means will be different from zero  which shows in TECH4. It is not clear to me if you want to plot as a function of the factors or the covariates. 


Thank you, that is helpful! For example, when the covariates are dummycoded demographic variables, I was going to plot a separate predicted probability for each covariate category, at low, mean, and high values of the factor of most interest, all other factors held constant at their mean. So for that reason, I wanted the means of all the latent factors. So I believe you've provided the info that I need to do that, since I'll take the means in Tech4. Thanks again. 

Back to top 

