Hi everyone,
I am modelling data from different studies together. Studies are grouped into references and so I have assigned the random effect of (1 | reference).
The response variable is calculated by dividing y by actors and sample days to get y per actor per day.
So the model could be:
brm( yActorDay ~ a + b + (1 | reference), family = lognormal() )
I am trying to change the weight of a study in the model based on the number of days that were sampled, and the number of actors they surveyed. I.e., so that studies that sampled a higher number of actors, and/or sampled them for longer periods of time were considered in the model to have ‘better’ estimate of the response variable.
I’ve seen that both offset() and weights can be used in brms(), but i’m not sure which/either would achieve the above?
Or indeed whether this is appropriate or necessary in a bayesian framework.