Run a Multivariate Bayesian Meta-Analysis in Stan
Usage
bayesma_mv(
data,
studyvar,
mean_ctrl_1,
mean_int_1,
sd_ctrl_1,
sd_int_1,
n_ctrl_1,
n_int_1,
mean_ctrl_2,
mean_int_2,
sd_ctrl_2,
sd_int_2,
n_ctrl_2,
n_int_2,
outcome_labels = c("outcome_1", "outcome_2"),
likelihood = c("gaussian"),
stage = c("two_stage", "one_stage"),
rho_within = 0.5,
mu_prior = NULL,
tau_prior = NULL,
rho_between_prior = NULL,
custom_model = NULL,
custom_data = NULL,
chains = 4,
iter_warmup = 1000,
iter_sampling = 1000,
adapt_delta = 0.95,
seed = 1234,
...
)Arguments
- data
A data frame with one row per study containing arm-level data for both outcomes.
- studyvar
Column for study identifiers (unquoted).
- mean_ctrl_1, mean_int_1
Columns for control and intervention means for outcome 1 (unquoted).
- sd_ctrl_1, sd_int_1
Columns for control and intervention SDs for outcome 1 (unquoted).
- n_ctrl_1, n_int_1
Columns for control and intervention sample sizes for outcome 1 (unquoted).
- mean_ctrl_2, mean_int_2
Columns for control and intervention means for outcome 2 (unquoted).
- sd_ctrl_2, sd_int_2
Columns for control and intervention SDs for outcome 2 (unquoted).
- n_ctrl_2, n_int_2
Columns for control and intervention sample sizes for outcome 2 (unquoted).
- outcome_labels
Character vector of length 2 with labels for the two outcomes. Default:
c("outcome_1", "outcome_2").- likelihood
Character. Currently only
"gaussian".- stage
Character.
"two_stage"(effect sizes computed then modelled) or"one_stage"(marginalised model).- rho_within
Numeric scalar in
[-1, 1]. Within-study correlation between the two outcomes, assumed known. Default:0.5.- mu_prior
Prior on pooled effects. Either a single prior (applied to both outcomes) or a named list with elements matching
outcome_labels.- tau_prior
Prior on between-study SDs. Either a single prior or a named list.
- rho_between_prior
Prior on the between-study correlation. Default:
uniform(-1, 1).- custom_model
Optional character scalar containing complete Stan code to override the generated program.
- custom_data
Optional named list merged into the Stan data list.
- chains, iter_warmup, iter_sampling, adapt_delta, seed
MCMC settings.
- ...
Passed to
cmdstanr::sample().