Hello. I am trying to run a linear mixed model on a small dataset where students read short reading passages and the spacing of the text varied between three possible settings. In the dataset, I have the reading speeds for their fastest and slowest reading speeds for each student along with the spacing variation (wide, normal, narrow). I also have some other factors I wanted to test like average reading speed across the whole test (there were other types of format we were testing earlier in the test). I'd like to know whether certain spacing variations benefitted reading speed in a broad sense and then individuals. I am trying to run mixed models with students as a random intercept and am having a lot of issues. I am often getting warnings that make me think I might be doing something wrong:
It's often one or both of these. The Hessian message is always when I try to add students as a random intercept.
|
Warnings
|
|
glmm: The final Hessian matrix is not positive definite although all convergence criteria are satisfied. The procedure continues despite this warning. Subsequent results produced are based on the last iteration. Validity of the model fit is uncertain.
|
|
Warnings
|
|
Iteration was terminated but convergence has not been achieved. The MIXED procedure continues despite this warning. Subsequent results produced are based on the last iteration. Validity of the model fit is uncertain.
|
|
The final Hessian matrix is not positive definite although all convergence criteria are satisfied. The MIXED procedure continues despite this warning. Validity of subsequent results cannot be ascertained.
|
I have found that sometimes removing the students as a random intercept removes this message depending on how many other variables I add to the model. Perhaps that is the issue, but I also worry I am just making an error with my syntax. I am also seeing a lot of blanks on my estimates of fixed effects for these. I just am not sure where to begin to figure out where I could be going wrong. Here is an example of some syntax:
MIXED T1SpacingWPM BY SpacingType FontType WITH CmpPerT1 AvgWPMT1
/CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0,
ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
/FIXED=SpacingType FontType SpacingType*FontType CmpPerT1 AvgWPMT1 | SSTYPE(3)
/METHOD=REML
/PRINT=SOLUTION
/RANDOM=INTERCEPT | SUBJECT(StudentID) COVTYPE(VC).
Adding students here throws the error. If I remove it, it goes away, but I am also still left with a lot of blanks on my fixed effects table for some of the interactions. I'm just not sure what I am doing wrong or if it is the data. Any ideas?
------------------------------
Stacy Adams
------------------------------