Attachment | Size |
---|---|
Screen Shot 2018-11-05 at 1.40.31 PM.png | 328.73 KB |
Growth models with TICs.R | 4.45 KB |
Hi everyone,
I am using OpenMx to fit a growth model with 2 continuous TICs. The results looked well; however, I noticed that the biases of path coefficients of the 2nd TIC were systematically larger than those of the 1st TIC in simulation studies. I am not sure if such patterns resulted from data generation, or from model misspecification, or from optimizer I used (the default one). I am attaching the script for one replication and the path diagram (with 6 repeated measures) of the model, any advice would be appreciated.
How much bias are we talking here? Could you provide a table of bias, MSE, etc. for the parameter esitmates of interest from your simulation? I don't see any reason why the regression coefficients would be more biased for one TIC versus the other.