Hello all,
I have a theoretical question. I am trying to run SEM, but one of my latent variables have 24 items as an indicator, and whenever I use it, the fit of the model goes down. So I thought of using it as a single indicator given that I know its reliability. However, there are some measures of which I do not know the reliability so I cannot use them as single indicators. Moreover, using multiple indicators for these variables would improve the degrees of freedom of the model.
So I was thinking if it is okay to use single and multiple items latent variables in the same model, or there is some theoretical reason for not doing it.
I looked around and couldn't find methodological papers explaining this.
Thank you.