|
|
Performing factor analysis before LPA |
|
Message/Author |
|
Sam Daneils posted on Thursday, September 24, 2020 - 3:38 am
|
|
|
Hi, I have a set of continuous indicators that are highly skewed. I performed LPA by keeping those indicators separate. One feedback I got from my committee member is to perform a factor analysis before adding those indicators separately in the LPA modeling. I'm a little bit hesitant to do that because I won't get good factor analysis results due to the skewness nature of my indicators. I also remember seeing something on this forum that performing factor analysis before LPA is not suggested. My questions are: 1. Is it appropriate to do factor analysis before LPA? 2. Could you please share some reference if there is any publication/documents on this topic? I really appreciate your help! |
|
|
Having skewed variables does not invalidate LPA. In fact, mixtures generate non-normal data. The only issue is if you have strong floor or ceiling effects. I am not sure factor analysis helps here. You could instead try to model correctly avoiding the usual linear assumption, for instance declare the variables as Categorical if they don't have many categories. I don't know that there is literature directly addressing this. You could ask on SEMNET. |
|
Back to top |
|
|