0
$\begingroup$

I want to perform a model with multiple independent variables and 2 random effects to see their effects on clutch size.

This is a plot of my response variable.

This is how my data looks like

My model

Clutch ~ temperature + I(temperature ^2) + rain + I(rain^2) + femalesize + FemaleAge + I(FemaleAge^2) + (1|YEAR) + (1|IDFemale)

I've seen a lot of ways to deal with this kin of distribution but i'm not sure which one is the best approch.

  1. linear mixed-effects model
  2. glmm with poisson distribution
  3. glmm with truncated Poisson distribution
  4. cumulative link mixed model
  5. tranform clutch size into binomial data (0 = clutch of one, 1 = clutch of 2 or 3)

Results between clmm and lmm are pretty similar. But glmm with poisson distribution are really different.

PS: when I check the diagnostic plot with DHARMa, poisson glmm looks problematic.

Improve this question
$\endgroup$
3
  • $\begingroup$ 1. A distribution over small integers like {1,2,3} (such as avian clutch sizes) could not reasonably be called "Gaussian". Mention of it seems to be a distraction from the issues at hand; I'd suggest removing it from your title. 2. Naturally an ordinary Poisson with mean somewhere in the region of 2 (and so with substantial probability on 0) is unlikely to be a suitable model for a variable that (we know a priori) won't have any zeros in it. $\endgroup$
    – Glen_b
    Commented Sep 30, 2020 at 7:16
  • $\begingroup$ Thanks for the awnser so if I understand correctly, it would be better to use either clmm (option 4) or logistic regression(5)? $\endgroup$
    – Ian.T
    Commented Oct 2, 2020 at 13:34
  • $\begingroup$ My comment mainly an attempt to get a more focused question by clarifying some issues with its framing and premises. I'm probably not the best person to answer it. $\endgroup$
    – Glen_b
    Commented Oct 3, 2020 at 2:56

0

Reset to default

Your Answer

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