1

I have some questions regarding a 3-way interaction-effect in linear mixed model analyses in R with the lme4 package.

My outcome measure is "MMSE" which is a cognitive test. I'm mainly interested in the decline in MMSE over time (time variable = "fuclin") by group ("ATgroups" 1,2,3,4) and cholesterol ("BL_CHOL" 0,1). These analyses are also corrected for several covariates (age_cent, GENDER, EDUC and APOE_dich). Dataset is called "ib".

This is what my model looks like:

summary(lmer(MMSE~ apoe_dich + EDUC + GENDER + age_cent + numfactors + fuclin* BL_CHOL*ATgroups + (1 + fuclin|ID),ib))

And this is what my output look like:

Output

So in this output, it always takes the lowest group of the interaction (ATgroups=1 and BL_CHOL=0) as a reference group. But if I add the interaction to the model using a colon (:) to the model, I get a different output. So my code would be:

summary(lmer(MMSE~ apoe_dich + EDUC + GENDER + age_cent + numfactors + fuclin: BL_CHOL:ATgroups + (1 + fuclin|ID),ib))

Then my output looks like this:

Output2

In this case, I'm not sure from which comparison the p-values are generated. And how can I get contrasts similar to the ones in output 1, but with the comparisons I want (instead of just comparing to the lowest category)?

Thanks!

Isabelle Bos
  • 105
  • 1
  • 7
  • Perhaps you could set the reference levels of the variables before you run the analysis. [Here](https://stackoverflow.com/questions/3872070/how-to-force-r-to-use-a-specified-factor-level-as-reference-in-a-regression) is a good example. – jacobsg Nov 01 '17 at 17:15

0 Answers0