Hello all!
My question could have either a very easy or relatively complex answer, but I am currently lost and would appreciate some helpful insight.
I've been given a dataset with results from a binary-choice experiment in which a certain cue was introduced (and the side of introduction was randomized). In summary this is what I have:
>4 treatment groups
>4 replicate groups inside each treatment
>Cue input side (2 sides)
> two interdependent response variables:
> proportion of responsive inds (PR)
> proportion of inds that responded correctly (PC)
PR = nr of inds that responded (R) / total nr of inds tested (T)
PC = nr of inds that responded correctly (C) / nr of inds that responded (R)
Since I want to compare both response variables (response and choice), between treatments, I was planning on applying a Binomial GLM (possibly GLMM if test side or replicate could have any effect, and would add them as random effect).
Shouldn't I apply some kind of hierarchical test, since my response variables are directly dependent of each other?
(from the animals that responded, some exhibited a correct choice and others a wrong one).
Perhaps GLMMM? (but is it a rabbit hole?).
Thank you in advance!
Hi Francisco,
I think that the type of analysis depend on questions which you want to answer. What are your hypotheses?
I assume that you have raw data. In that case, instead of 2 bivariate response variables, you can create one response variable with three categories (did not respond/ responded wrongly / responded correctly) and model it as multinomial response (similar models as for binomial).
Best regards,
Ondrej
Hello Francisco, I suggest that you seek the support of epistata.org They are an independent agency for clinical research, epidemiology, and medical statistics. They will help you. All the best
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R