|
|
MLR to test moderation models? |
|
Message/Author |
|
Jake Lant posted on Friday, March 02, 2012 - 5:44 am
|
|
|
Hello friends, I have a question about the method I am using—I used MLR (M-estimation with robust SEs, correct?) to test a moderation model. DV and proposed moderator are both continuos, and the third variable is categorical (gender); data "symptoms": non normal distribution, outliers, missing data (around 1%). n = 467. MLR seemed like the best way to go given the situation I have here, but after reading 1987 Rousseeuw and Leroy's book (Robust regression and outlier detection), I am questioning my choice. There are way too many methods, but do you think I made the best choice, based on the methods available on Mplus? Or is there a better method that could perhaps prove to be a more powerful test? Thanks in advance, JL |
|
|
MLR is a good estimator choice. It is robust to non-normality. If you are worried about outliers, use the LOGLIKELIHOOD option of the SAVEDATA command. |
|
Back to top |
|
|