I have never posted here before, so apologies if I do not follow the correct format.

My experiment design is I have 12 reps each of 4 different species of plant which I experimented on in 2 blocks, so 6 of each of the 4 species in Block 1, 6 x 4 in Block 2. I used the plants to rear some bugs, and am now looking at various response variables of the bug (body length, wing morph etc) to see if there's any difference due to their host plants. Bugs were also able to breed during the experiment, so there are different numbers of bug on each plant - I've standardised this using a crowding measure (no. bugs / g plant biomass). There were 1138 bugs in total on the 48 plants. There is an individual measurement for each bug, but numerous bugs per plant.

My model at the moment looks like this:

model<-lme(fixed=log(length)~Species*Crowding*Sex, random= ~1|Block/PlantID, method="ML",data=inds)

Biologically speaking, I'm not at all interested in the effects of block or plant ID, but have put them in as random effects to account for pseudoreplication. Although not explicitly designed into the experiment, I am interested in the effect of crowding, particularly if the response to crowding is different depending on the host plant species and / or if sexes respond differently to crowding levels.

However, the Crowding and Plant ID variables are autocorrelated (as each bug on plant 1 will have the same crowding value, as will each bug on plant 2 etc) so I am accounting for this variation twice. As I understand it, I can't remove Plant ID due to the issue of pseudoreplication, so how can I test for the effect of crowding and its interactions whilst removing autocorrelation from my model?

Any help would be greatly appreciated, thank you.


Your Answer

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

Browse other questions tagged or ask your own question.