Hi, I was just interested to know if anyone had any insight into a model I am currently trying to run. In brief I am interested in seeing how the mean and variance in behaviour (in this case animals) might be correlated to reproductive success. In my example I have multiple measurements of an individuals behaviour within a given year but only a singular measure of their reproductive success.

A mock dataset for what I am looking at is below.

|Column 1 | Column 2 | Column 3 | Column 4 | E | F|

|Ind_ID | Foraging_time | Temperature | Year | Breed_success | Sub|

|Animal_1 | 40 | 5 | 2023 | 1 | T|

|Animal_1 | 20 | 5 | 2023 | NA | F|

|Animal_1 | 30 | 7 | 2023 | NA | F|

|Animal_1 | 25 | 7 | 2023 | NA | F|

|Animal_2 | 10 | 5 | 2023 | 0 | T|

|Animal_2 | 15 | 5 | 2023 | NA | F|

|Animal_2 | 10 | 7 | 2023 | NA | F|

|Animal_2 | 15 | 7 | 2023 | NA | F|

|Animal_3 | 60 | 5 | 2023 | 1 | T|

|Animal_3 | 30 | 5 | 2023 | NA | F|

|Animal_3 | 20 | 7 | 2023 | NA | F|

|Animal_3 | 40 | 7 | 2023 | NA | F|

I am attempting to run this model in brms with mean and variance components for behaviour and just mean for the reproductive success. I would like to run this as a bivariate model so that I can account for âfixedâ effects on both response variables. What I am wondering is if it is possible to set a prior so that the within-individual variance of a random effect can be fixed at 0 whilst still allowing the between-individual variance to be calculated in order to quantify the correlation between foraging and reproductive success. I have attempted to use the constant(0.0001) setting when setting the prior for the breed_success model, but I think it is removing all variance and therefore not allowing the calculation of the correlation between the two measures (when the model runs, the variance is set to 0 but the correlations are also remarkably close to 0 for me). I have previously run a similar model in MCMCglmm before that allows the within- and among-group variances to be set differently (hence the within- can be fixed to 0) but do not know if there is a similar way of doing this in brms. An example of the type of prior I am thinking of in MCMCglmm can be found on PAGE 17 of this document https://tomhouslay.files.wordpress.com/2017/02/indivvar_plasticity_tutorial_mcmcglmm1.pdf. If anyone knows if this is possible for a prior setting to allow there to be among-individual variance without any within-individual that would be greatly appreciated.

The model structure currently would look like this

m.f.t â bf(Foraging_time ~ Temperature + Year + (1|q|Ind_ID)) + gaussian()

m.b.s â bf(Breed_success | subset(Sub) ~ Year + (1|q|Ind_ID)) + bernoulli()

model â brm(m.f.t + m.b.s,

data = data,

prior = prior,

warmup = 1000,

iter = 5000,

thin=4,

chains = 4,

seed = 12345)

Although, I understand I only have single measures of an individuals Reproductive success and therefore setting a random effect is odd, it is required to calculate the correlation between the two responses.

Any help on this topic would be greatly appreciated and I look forward to hearing from anyone!

Cheers,

Freddie