Hi everyone,
I’m interested in doing meta-analysis using brms, and wanted to know if there was any way to calculate I2 for improved interpretability when describing the results.
My model is the following: where I pre-calculated yi and seTE using escalc. I had to use StudyDesign as a fixed effect due to too few studies and non-convergence despite varying priors. I’m wondering if it is possible to get weights to back-calculate Q and then use that to get I2 from each draw.
Thanks so much!
brm.count ← brm(yi | se(seTE, sigma = FALSE) ~ 1 + StudyDesign + (1 | AuthorYear), data = df_count_es,
prior = c(prior(cauchy(0, 0.5), class = Intercept),
prior(normal(0, 1), class = sd, group = AuthorYear)),
iter = 4000, warmup = 1000, cores = 4, chains = 4, seed = 1,
sample_prior = TRUE,
control = list(adapt_delta = 0.99, max_treedepth = 15))