Attachment | Size |
---|---|
bivariate moderation syntax.R [6] | 10.47 KB |
result1.txt [7] | 3.99 KB |
result2.txt [8] | 4.01 KB |
Hello everyone!
I am trying running a bivariate moderation model put forward by Purcell with binary moderator(smoke or not) and binary outcome(T2DM or not). I have running the model twice with the same syntax and the same data without any change, However, the results are different, not only the point estimate but also the -2LL . My syntax and results are in the supplementary materials.
I use the "NPSOL" optimizer and mxTryHardOrdinal() with extraTries=30. I also used a custom compute plan with Nelder-Mead advised by professor Robert. "set.seed(100)" is included in the model.
the output of "mxVersion()"
OpenMx version: 2.17.2 [GIT v2.17.2]
R version: R version 3.5.3 (2019-03-11)
Platform: x86_64-w64-mingw32
Default optimizer: CSOLNP
NPSOL-enabled?: Yes
OpenMP-enabled?: No
There is a warning in my result: "The model does not satisfy the first-order optimality conditions to the required accuracy..." From the forums, I know that the warning is quite common for categorical variables.
My biggest puzzle is that the results of the twice running are different, which means the result is unreliable.
I'll be really grateful if anyone can give some suggestions!