I am new to SEM and have some methodological questions. Here is the model I would like run (LV = latent variable and OV = observed variables). enter image description here

  1. First, from the example I saw on the web, gender, age etc. are often "covariates". However, due to the litterature, I have specific hypotheses that these observed variables have direct relationship on LV_A, LV_B and even LV_C. I also have the hypothesis that age has a direct relationship on challenging behavior. For this reason, I have simple arrows. Is that correct or should I change that ?

  2. Given that "comorbidities" is a covariate, should I have double arrows between comorbities and all latent varaibles?

  3. I want to pre-register the study but I am not sure that the model will fit (because I might have fewer participants than expected). In case it does not fit, I plan to (i) run two models instead of one (one for the blue variables and one for the green variables, even if I will loose some information). (ii) I also plan to remove some variables such as comorbidities. (iii) Finally, I plan to reduced some of the observed variables (actually, for LV_A, I have 7 variables, that be reassemble in 2 variables that can finally be reassemble in LV_A; So I was thinking about taking the 2 variables instead of the 7). What do you think about pre-register that?

Thank you for your help



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.