If just started analysing data I got from around 250 questionnaires. I've built a model in AMOS but I can't get it to match what is going on with my linear regression! In my LR all variables are seen to independently contribute to the DV, but these pathways aren't significant in my SEM model Would this be because I used global questionnaire scores for my L.R whilst I built my model with latent variables were all questionnaire data was used but in their relative component scores rather than one single global score? Am I right in thinking this would bring variance into the data?
If so, would it be best to build a new model wherein I just use single global scores as observed variables or leave my model as it is with latent variables made from sub scores? I get good fit for both models, but for the observed model - I get a really high RMSEA due to the low DFs!
Would this be because I used global questionnaire scores for my L.R whilst I built my model with latent variables were all questionnaire data was used but in their relative component scores rather than one single global score? Am I right in thinking this would bring variance into the data?
Posting Permissions
Reply With Quote