0

I am running a 3-way interaction which includes a binary variable (race) and 2 continuous variables (age and dep) covarying on sex, SES, and drug_use. See code below

m1  <-  lme(Anx ~ sex + SES + drug_use + race + age + dep + race:age + age:dep + dep:race + age:race:dep,
            random = ~1|ID, 
            data = df, method = "REML", na.action = na.omit)
summary(m1)
intervals(m1, 0.95)

To probe significant interactions, I stratified by race. However, I was told that I didn't have to run separate race stratified analysis to get the race-specific estimates, but that these estimates can be directly derived from the model with interactions. Does anyone have any ideas on how to do this?

Ronak Shah
  • 377,200
  • 20
  • 156
  • 213
D. Fowler
  • 601
  • 3
  • 7
  • Can you not compare nested models using the `anova()` function? – sjp Mar 16 '21 at 04:25
  • try the `emmeans` package? – Ben Bolker Mar 20 '21 at 01:01
  • @BenBolker, I did that and that is part of what I need. However, I also want to calculate the simple slopes. I tried doing this using the simple slopes function in reghelper, although I got some error messages doing that. I submitted a separate post about this. – D. Fowler Mar 20 '21 at 01:20

0 Answers0