Message/Author 


I am running a moderation of one observed continous variable Z on the path between latent variable X on latent variable Y. The continuous variable Z is a stated percentage of usage of self service technology (SST) rated from 0% to 100% of the time. Since this variable does not follow a normal distribution, I thought it would be best to do a log transformation (base10) of Z prior to running the interaction. This leads me to two questions: 1) Can I write the code for the transformation in the command syntax or do I have to do it in the raw data? If possible in the syntax, where would I find sample code? 2) Would the log transformation cause problems with the estimation? ie: give improper estimates. Thanks, 


You can use the DEFINE command to do the log transformation. The log transformation would not cause estimation problems. 


How well does Mplus cope with violations of linearity? Using MLR with nonnormal and skewed ordinal data. Two of my IVs have high floor effects. My model fit remains poor and unchanged even after removing nonsignificant paths. I'm trying to think of what could be biasing the model. 


Skewed IV distribution does not necessarily cause problems. It is mostly relevant for DVs. If the DVs are treated as ordinal, skewness is not an issue. 

Back to top 