My design goes like this: I have 1 treatment and one control, organized in 3 blocks, each have 1 site of control and 1 site of treatment, each site have 2 subsites, and I sampled 6 quadrats per subsites: 3 under tree canopy, 3 outside tree canopy. This is a repeated measure experiment, with the same sampling method every year. I am interested in comparing treatment vs control over time.

My model goes like this: lmer(Y ~ Time * Treatment + (1 | Block) + (1 | Site / Subsite)). But I am not sure how to classify the canopy component. Should I include it as a random effect nested in subsite as (1 | Site / Subsite / Canopy) or on its own as (1 | Canopy) or as a fixed effect? (Or can I not include it in the model as I am not interested in it??)


I would say use it as a fixed effect, because:

  1. With only two levels, you don't have enough data for adding it as a random effect.
  2. Don't know about your particular case, but it sounds quite likely that the Canopy will have effect on the response variable. Is it the case? In such case it makes perfect sense to include it in the model as a fixed effect. If the canopy has any significant effect on the Y, then you should include it in the model since it can help you to explain more variability by the other covariates as well.
  • $\begingroup$ Thanks, that makes a lot of sense! $\endgroup$ – user254059 Jul 19 '19 at 18:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.