Keith Tan posted on Wednesday, November 25, 2015 - 9:42 pm
I am trying to understand an SEM analysis from my colleague. I've asked her to explain it to me, but I find it confusing and I suspect something wrong with her understanding.
For me, the final model diagram seems very complicated and I am unable to clearly deduce what the research question might be from it or what the endogenous variables or variables of interest are. Unfortunately, she also cannot give me a clear answer on what these are.
She says that in specifying the model syntax, she tried out all possible relations. I asked if that meant all possible relations like 1 to 2, 1 to 3.. 2 to 3, 2 to 4 and also in the opposite direction etc. and she said yes.
It sounds very suspicious to me as with 7 variables, the possible number of identified/over-identified models will range in the hundreds if not thousands.
From what I have shared, could I ask your help to share possible scenarios of what my colleague actually did? Would you also call this an SEM with an exploratory approach?
Measurement model: 7 latent variables were calculated from the observed data, i've labeled them 1 to 7. they include constructs like efficacy, agency and practice etc. model fit indices are good (i am not including them here though)
Structural Model: Below is the syntax on the structural relations.
From your description I don't think the person has enough theory to do SEM. I wouldn't call that exploratory SEM, but rather not knowing enough about the subject matter behind the variables to do a meaningful analysis. You may want to ask on SEMNET.