Adam Blake posted on Friday, May 04, 2012 - 4:12 pm
I am attempting to fit a SEM with a categorical variable. I am using a series of dummy variables and wanted to test the overall effect of the categorical variable using a composite/formative variable. When trying to fit the effect of the categorical variable on multiple endogenous variable should I use a single composite variable with paths to each endogenous variables or create a composite variables for each endogenous variable?
Good question. The formative factor is really interpretable only relative to a certain DV, so you could in principle have one formative for each DV, where each formative has different weights for the dummies. So it depends on your substantive thinking.