I have been fitting a bayesian GLM using brms. The code works well but when I loop this over several data and make it a bit more complex, R encounters a fatal error and crashes. This seems to be happening to other people when using brms. I am now looking into other methods to code this in R (eg, stan_glm, bayesglm), but they don't seem to accept bounded uniform priors like brms does, which is what I need. Does anyone have suggestions? Thanks
code
N <- c(10, 12, 15, 20, 18)
success <- c(2, 4, 6, 8, 10)
x1 <- c(0, 0, 0, 0, 5)
x2 <- c(0.5, 0.5, 0.5, 0.5, 0.5)
data <- data.frame(N, success, x1, x2)
prior1 <- c(set_prior("uniform(-4,0)", class = "b", coef = "x1"),
set_prior("uniform(-4,0)", class = "b", coef = "x2"))
fit_bglm <- brm(success | trials(N) ~ -1 + x1 + x2,
data = data,
family=binomial(link = "log"),
prior=prior1, chains = 8, iter = 8000,
seed=154511)
link = logis not typically a good link function for the binomial distribution. 2) (a simplification tip) Why do you create a new variablex2equal to 1/2 of what the intercept would be instead of just using the intercept? You can set a prior on the intercept directly viaclass = "Intercept". 3) It would be better to give us an example which breaks, rather than an example which works; it's hard to debug code that appears to be working. $\endgroup$