I have a study design on two sites (each site was set up in a different way) and am trying to find the best approach for the analysis. I'm looking at survival of plant species, under 4 treatments, on 2 sites. Site 1 is blocked by treatment and site 2 is blocked by tree species. Before I realized the 2 sites were set up differently, I planned on running a nested ANOVA or a Mixed Model Regression but I'm confused on how to proceed as is. Coding (using R) in the block effect will just lead to co-linearity on Site 2 with my oaks, right? I suppose I could analyze the sites separately, but I want to know if there is a way to analyze them together. I uploaded a diagram of the study design. The study is balanced, there are 5 subplots/block. Appreciate any advice.

enter image description here

  • 2
    $\begingroup$ Go to the people who did this study and tell them "THIS is why you should consult the statistician EARLY!" $\endgroup$
    – Peter Flom
    Commented Jul 10 at 17:10
  • 2
    $\begingroup$ I did. They don't understand that the post-mortem for this kind of thing is difficult. I have made a check list that needs to be reviewed and answered for upcoming studies. $\endgroup$ Commented Jul 10 at 17:40

1 Answer 1


Here's what I came up with for this issue: In this particular design my blocks were close to one another in proximity and running the above regression showed low level residuals by block. So I dropped block and added site as a fixed factor to incorporate the limitations of the design. Running this way puts some limits on being able to extrapolate the study to a broader whole but it gets around the initial issues. Simple fix, hopefully it's helpful to someone in a similar situation.


Your Answer

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

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