0

I have 3 treatments (A, B, C) and 30 centers. Each center only administers one treatment to its participants. Example, center1 administers Treatment A to its participants, center2 administers Treatment B, and so on.

Data on outcome and demographics are collected from the participants.

I would like to find out the effect of the treatments on the outcome after controlling for demographics. So fixed effects are treatments and demographics.

Participants in the same center are dependent. Hence, would like to model center as 2nd level random intercept. Question: is it meaningful to model center as random intercept given that each center only administers one treatment?

Robert Long
  • 5,722
  • 5
  • 29
  • 50
Bernice Liting
  • 1

1 Answers1

0

Question: is it meaningful to model center as random intercept given that each center only administers one treatment?

Yes. The main point here is your previous two sentences:

Participants in the same center are dependent. Hence, would like to model center as 2nd level random intercept.

Exactly. We fit random intercepts for groups when we expect dependencies within those groups. That each center administers the same treatment is not important, except that it may lead to a loss in statistical power, so a sample size study should be conducted. This type of study design is used in cluster-randomised controlled trials.

Robert Long
  • 5,722
  • 5
  • 29
  • 50