Hi everyone, I would like to perform model comparison between 5 Hierarchical GAMMs using brms() and as part of it, I need to perform Leave One Out Cross Validation, which is performed by the loo() function in brms.
Below is all my code leading to my RStudio console crashing… I read that running the models and loo() in the R terminal while specifying only 1 core, tended to resolve the issue, but that has not worked for me either. I really need to be able to perform this analysis as it’s part of my dissertation.
Here is the code:
rm(list=ls(all=TRUE))
library(mgcv)
library(brms)
library(readxl)
library(cmdstanr)
library(bayesplot)
library(ggplot2)
library(marginaleffects)
library(tidyverse)
library(tidybayes)
library(loo)
library(performance)
# Load data
data <- read_excel("~/data.xlsx", sheet = "data1")
# Specify categorical variables as factors
data$ind <- as.factor(data$ind)
data$town <- as.factor(data$town)
fit.G <- brm(bf(prop.fruit ~ s(time.index, m=2, bs="tp") +
(1|ind) +
(1|town),
zi ~ s(time.index) +
(1|ind) +
(1|town)),
family = zero_inflated_beta(link = "logit"),
prior = NULL,
data=data,
chains=4,
iter=8000,
warmup=4000,
cores=4,
seed=1234,
backend="cmdstanr")
loo.G <- loo(fit.G, moment_match = TRUE)
After running this code, my RStudio experiences a fatal error and crashes (see image below). Please, I would really appreciate any recommendations, as I feel really close to finishing this analysis and only have this as the last obstacle in the way.
