Which statistical analysis is most appropriate to determine differences between 3 treatments and control for monthly and yearly variation?

Question

I have a 4-year dataset of zooplankton biomass that was sampled three times a summer (i.e. June, July, August) for all four years. There are 6 sites that were sampled in triplicate (took 3 reps): two sites in Treatment A, two sites in Treatment B, and two sites in Treatment C. There are 72 data points per zooplankton functional group once I average the reps for each site giving one measurement of biomass per site for a given month and year.

I want to see if there are differences in biomass between treatments controlling for year and monthly variation. What was suggested to me was the following model:

   Bbos.model <- lmer(log(Bos.Biomass+1) ~ Treatment + Year + 
                      (1|Site), data=mydata)

However, as I understand it, this treats the monthly sampling events as independent events which I don't think I can do. Is the correct analysis of this data a repeated measures analysis? Or would a mixed model approach be more appropriate?

Answer 1

The model that was suggested is actually a mixed model (the random part is composed of the real random term and 1|Site). Why don't you add Season as a factor if you want to control for it ?

It is not always easy to decide whether you add a variable as a fixed or as a random effect; this website http://bbolker.github.io/mixedmodels-misc/glmmFAQ.html and the book Mixed Effects Models and Extensions in Ecology with R could help you with this decision.