Mediation with brms, as_draws function

General
,
1

Dear all,

Sorry if this is somehow a simple question but I have been stuck on this for a while.
I am trying to do a simple mediation analysis. First I fitted a multilevel model in brms (“structural equation model”):

# Define structural equation model

path1 <- bf(bci ~ faith + cort + immunity + age + (1|group/ID))
path2 <- bf(immunity ~  faith + cort + age + (1|group/ID))
path3 <- bf(faitth ~ cort + age + (1|group/ID))
path4 <- bf(cort ~ age + (1|group/ID)) + skew_normal()

sem_immunity_faith <- path1 + path2 + path3 + path4

# Run brms

ncores = detectCores()
options(mc.cores = parallel::detectCores())

model_immunity_faith <-brm(sem_immunity_faith + set_rescor(FALSE),
                         data = metadata,
                         warmup = 50000, iter = 100000,
                         cores=ncores, chains=4, init=1000)

To do a mediation analysis I was trying to follow the instructions in this post:
https://discourse.mc-stan.org/t/mediation-question-using-brms/4260

But apparently the function posterior_samples is deprecated so I tried to do it with the new as_draws function, like so:

# Indirect effect of cort on faith meditated by immunity
post_immune_cort  <- as_draws_df(model_immunity_faith, variable = "b_immunity_cort")
post_immune_faith <- as_draws_df(model_immunity_faith, variable = "b_immunity_faith")

indirect_cort_faith <- post_immune_cort *post_immune_faith 

indirect_cort_faith <- as.data.frame(posterior::summarize_draws(indirect_cort_faith))

It seems to work but I am not sure if I am doing something wrong. Any feedback on this?

Thanks in advance.
Kind regards,
Hugo

2

Maybe this is a better way:

post_immune_cort  <- as_draws_matrix(model_immunity_faith, variable = "b_predimmunity_cort")
post_immune_faith <- as_draws_matrix(model_immunity_faith, variable = "b_predimmunity_faith")
indirect_cort_faith <- post_immune_cort *post_immune_faith 

indirect_cort_faith <- summarise_draws(indirect_cort_faith,"mean","median","sd",
						~quantile(indirect_cort_faith, probs = c(0.025, 0.975)),"rhat","ess_bulk","ess_tail")

Kind regards,
Hugo