I am trying to conduct a Mixed model multilevel analysis.

One of my hypothesis is, that daily Hindrance stressors (HS) will be related to negative affectivity in the evening (NA_EV).

We had 3 daily questionnaires over 14 consecutive days. We asked people about their affectivity in the morning (NA_AM) and in the Evening (NA_EV) and about their daily hindrance stressors in the afternoon (HS_PM).

Now I want to control for negative affect in the morning, in order to only see the change in negative affectivity during the day insted of a total value.

When I add Hindrance Stressors to my model, I get the following Warning: 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.

I do not get what they mean by it, especiallly because I am doing the exact same thing with challenge stressors and positive affectivity and there I am not getting this warning. If anyone could help me, I would really appreciate it!

My Syntax looks like this:

MIXED NA_EV

 /CRITERIA=DFMETHOD(SATTERTHWAITE) CIN(95) MXITER(100) MXSTEP(10) SCORING(1) 

  SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)   

 /FIXED=| SSTYPE(3)

 /METHOD=ML

 /PRINT=SOLUTION TESTCOV

 /RANDOM=INTERCEPT | SUBJECT(participantnr) COVTYPE(ID).

MIXED NA_EV WITH NA_AM

 /CRITERIA=DFMETHOD(SATTERTHWAITE) CIN(95) MXITER(100) MXSTEP(10) SCORING(1) 

  SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)   

 /FIXED=NA_AM | SSTYPE(3)

 /METHOD=ML

 /PRINT=SOLUTION TESTCOV

 /RANDOM=INTERCEPT | SUBJECT(participantnr) COVTYPE(VC).

MIXED NA_EV WITH NA_AM

 /CRITERIA=DFMETHOD(SATTERTHWAITE) CIN(95) MXITER(100) MXSTEP(10) SCORING(1) 

  SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)   

 /FIXED=NA_AM | SSTYPE(3)

 /METHOD=ML

 /PRINT=SOLUTION TESTCOV

 /RANDOM=INTERCEPT NA_AM | SUBJECT(participantnr) COVTYPE(UN).

MIXED NA_EV WITH NA_AM HS_PM

 /CRITERIA=DFMETHOD(SATTERTHWAITE) CIN(95) MXITER(100) MXSTEP(10) SCORING(1) 

  SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)   

 /FIXED=NA_AM HS_PM | SSTYPE(3)

 /METHOD=ML

 /PRINT=SOLUTION TESTCOV

 /RANDOM=INTERCEPT NA_AM | SUBJECT(participantnr) COVTYPE(UN).

Well, you are doing something different when you add that variable. You could have a collinearity or small cell problem.

I suggest two things to try. Up the maximum number of iterations, currently set to 100, and, perhaps more important, try using REML rather than ML estimation.

Tank you for your help!

Unfortunately that doesn't work. The warning stays the same.

I am woundering, whether it is possible to still use the data SPSS produces, because the only thing that seem off is that in the covariance table the last row of Intercept+NA_AM does not produce any vlalues. And I am not needing these to answer my research question anyways.

Or is the whole output invalid after this warning?

I can't answer that question, but I suggest using the HISTORY keyword on the PRINT subcommand to see how the results are changing with the iteration. It would also be good to try simplifying your model in various ways to see what feature causes convergence to fail.