0

I'd like to present you with a problem. I cannot replicate in R's lme (nlme package) a repeated measures model I built in SPSS. I hope someone is able to solve this conundrum. See below for the steps I took.

Thanks!

My SPSS model is the following:

MIXED DV BY DX Visit WITH Cov
  /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) 
  /FIXED=Diagnosis Visit Diagnosis*Visit Cov | SSTYPE(3) 
  /METHOD=REML 
  /REPEATED=Visit | SUBJECT(ID) COVTYPE(UN).

My R model is the following:

model = lme(DV ~ Diagnosis * Visit + Cov, random = ~1 | ID, data = dataSet, correlation = corSymm(form = ~1|ID))

I tried many things, including modeling slope and including nested random effects.

The interesting thing is that if I model repeated measures as follows in SPSS (so with RANDOM instead of REPEATED):

MIXED DV BY DX Visit WITH Cov
  /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) 
  /FIXED=Diagnosis Visit Diagnosis*Visit Cov | SSTYPE(3) 
  /METHOD=REML 
  /RANDOM=Visit | SUBJECT(ID) COVTYPE(UN).

it gives me the same results as my first SPSS model, but it says that the model did not converge.

I'd really appreciate any input. Thanks!

best, Hans van der Horn

Hans van der Horn
  • 1
  • Some additional info: most factors/interactions are (almost) similar in R's lme, however, the main effects of diagnosis is way different from SPSS's output. – Hans van der Horn Mar 07 '23 at 15:58

0 Answers0