Adam Blake posted on Friday, May 04, 2012 - 4:19 pm
I have been trying to replicate the results of simple linear model with one categorical predictor performed in R with the results of model in M-plus using dummy variables with paths to a composite variable with a path to a endogenous observed variable. I realize these methods differ but my p-values seem to be much smaller and my R squared value seems to be much larger in my R model. This is true even when I fit the linear model in R using the same dummy variables. I am curious to know what might be causing these differences?
It sounds like you are comparing the formative factor Model 1 versus the regular regression Model 3 on slide 244 of our Topic 1 short course handout on our web site. They are equivalent models so should give the same result.