0

I'm trying to find variance in infectivity trait of animals in different herds. Each herds contains a fixed number of offspring from 5 different sires.

Example of data:

Herd S C DeltaT I sire1 I1 sire2 I2 sire3 I3 sire4 I4 sire5 I5
1 20 0 14 1 13 0 26 0 46 0 71 0 91 1
1 1 0 14 5 13 1 26 0 46 2 71 1 91 1
18 4 0 14 13 2 5 52 4 84 2 87 2 98 0
19 11 3 14 27 2 6 13 7 18 3 46 5 85 6

Herd is the herdname. S is the number of susceptible animals in the herd, C is the number of cases in the time interval. DeltaT is the time interval length. Sire# is the ID of the sire in the Herd. I# is the number of infected Ofspring of the corresponding Sire#.

This means that a sireID "13" in the first two rows in the column sire1. Refers to the same sire as the "13" in sire2 of the last row. To include these 5 sires into one random effect in a glmer of lme4 is getting me in trouble.

I tried:

glmer(data = GLMM_Data,
              cbind(C, S-C) ~  (1 | Herd) + (1| (I1 | sire1) + (I2 | sire2) + (I3 | sire3) + (I4 | sire4) + (I5 | sire5)), 
              offset = log(GLMM_Data$I/nherds * GLMM_Data$DeltaT),
              family = binomial(link="cloglog"))

This gave errors. So any help on combining these 10 columns in a single random factor would be more than welcome. Thanks in advance.

p.s. I know my offset, family and the left side of the formula are working since the analysis of susceptibility is working

Tim Bosman
  • 1
  • 3
  • I think you might be looking for *multimembership* models, which you can do via an extra layer with `lme4`: https://bbolker.github.io/mixedmodels-misc/notes/multimember.html, *or* with the `brms` package (but that starts you down the Bayesian rabbit hole). The linked notes have a wrapper function that works for `lmer`, but not one for `glmer`. I could write one for `glmer` ... – Ben Bolker Jun 17 '21 at 15:21
  • Thank you very much. I think I can fix the problem now. If I solved the issue, I will post the code below. – Tim Bosman Jun 18 '21 at 10:43

0 Answers0