9 Final LMM: Inputs

9.1 Final Model

Based on the previous model outputs, we decide to remove the interaction term of Group * Time from the model and refit the model keeping other fixed effects as main effects. We still include the random intercept of patient.


9.2 SPSS: Menu

  • Fixed

    • Remove the interaction term of Group * Time from Fixed Effects dialog box (in Model section)

    • Select the interaction term from the previous model and remove the item from the model clicking Remove.

    • This means that the model only includes the main effect.

  • Random

    • No change is required for the Random Effects dialog box.

  • Estimation

    • Select Method as Restricted Maximum Likelihood (REML)

  • Statistics

    • No change is required

  • EM Means

    • Estimate the marginal means of fitted model.

    • Dispaly means for all main effects

    • We are not interested to compare the Time effects. If we do, we can select appropriate strategies to compare the means controlling for type 1 error

No other changes required for fitting the revised model. Click OK from the main dialog box.

We need some other changes for model diagnostics, but first we should check that the revised model converges.

Once we are happy with the convergence, we will conduct model diagnostics to check assumptions. See the next section.

Here, we present all steps together.


9.3 SPSS: Script

MIXED SBP BY Sex Group WITH SBP0 Age BMI Time 
  /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 = Sex Group SBP0 Age BMI Time | SSTYPE(3) 
  /METHOD = REML 
  /PRINT = G R 
  /RANDOM = INTERCEPT | SUBJECT(ID) COVTYPE(VC) 
  /EMMEANS = TABLES(OVERALL) 
  /EMMEANS = TABLES(Sex) 
  /EMMEANS = TABLES(Group)