Posterior summary of a component ML-NMR fit

Summarizes the posterior draws of a cmlnmr() fit: the component effects beta, the component by effect-modifier interactions gamma, the study baselines mu, the heterogeneity tau, and any other sampled block, with the usual convergence quantities alongside.

Use this rather than reaching into fit$fit. That object is whatever the sampler backend returned: an S4 stanfit under backend = "rstan" and an R6 object under backend = "cmdstanr". The two share no accessors, so fit$fit$summary(...) works on one and fails on the other. This function works on both and returns the same columns either way.

Usage

posterior_summary(x, variables = NULL, ...)

Arguments

x

A cpaic_mlnmr fit from cmlnmr().

variables

Character vector of Stan variable names to summarize, for example "tau" or c("beta", "gamma"). Naming a block returns one row per element of it. The default summarizes every sampled block the fit has.

...

Further summary functions passed to posterior::summarise_draws(), for example "quantile2" or a function. With none given, the default set is returned.

Value

A data frame with one row per scalar parameter. With the default summaries the columns are variable, mean, median, sd, mad, q5, q95, rhat, ess_bulk, and ess_tail. Because both backends are summarized through the same code, rhat, ess_bulk, and ess_tail are the same quantities whichever engine produced the fit.

See also

cmlnmr() for the fit, relative_effects() and component_effects() for effects on the outcome scale rather than the parameters themselves, and redact_fit(), which strips the draws.

Examples

if (FALSE) { # \dontrun{
fit <- cmlnmr(ipd, agd, effect_modifiers = "x1", inactive = "Placebo")
posterior_summary(fit, "tau")
min(posterior_summary(fit)$ess_bulk)
} # }