I'm working with daily diary data that include a categorical variable indicating which of 4 possible events happened that day. I dummy-coded this variable and entered it into my multilevel model. However, I'm unsure what to do at level 2. I would typically enter a person-mean variable into level 2 and a person-mean-centered variable into level 1 for continuous variables. Do I need to do something similar for this variable (e.g., enter the person-specific percent of days with a given event)? And if so, do I enter level 2 variables for all 4 events, or only the 3 events entered into the model?

  • $\begingroup$ Good question. I am curious as to what others will have to say about this. My initial thought would be to include just the group means for the three indicators in the model. They completely cover all cases for the time-varying covariate, and their means would represent the % of cases for an individual who was in category 1 (relative to 0), category 2 (relative to 0), and category 3 (relative to 0). That being said, I can see an argument for also including the mean for category 0. Have you tried running a model with 3 and 4 group means? $\endgroup$
    – Erik Ruzek
    Aug 1, 2019 at 22:42


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.