summary.brma creates summary tables for a
brma object. For RoBMA objects, inclusion summaries are printed before
parameter estimates. Random-effect inclusion rows are included in
Component Inclusion, immediately before Publication Bias when present.
Their labels prefix the corresponding SD quantity with Heterogeneity:, retaining
component names where needed. Data-frame exports use the same layout.
Arguments
- object
a fitted brma object
- probs
quantiles of the posterior samples to be displayed. Defaults to
c(.025, .50, .975)- include_mcmc_diagnostics
whether to include MCMC diagnostics in the output. Defaults to
TRUE.- standardized_coefficients
whether to show standardized formula coefficients. Defaults to
FALSE. When set toTRUE, meta-regression coefficients and random-component summaries are returned on the standardized scale on which prior distributions are specified by default (i.e.,standardize_continuous_predictors = TRUE). A random-effect row whose original-scale value the fitted random block cannot represent (for example the intercept of a block with factor-specific slopes of a standardized predictor) isNAin the default original-scale table, with a footnote; its value on the fitted scale is shown withstandardized_coefficients = TRUE.- conditional
whether to include conditional estimates for RoBMA product-space objects. Defaults to
FALSE.- logBF
whether to show inclusion Bayes factors on the log scale. Defaults to
FALSE.- BF01
whether to show inverse inclusion Bayes factors. Defaults to
FALSE.- ...
unused additional arguments. Supplied arguments trigger a warning.
- x
a
summary.brmaor fittedbrmaobject.
Value
A list of class summary.brma with model name, optional RoBMA
inclusion tables, common estimates, moderator estimates, scale estimates,
publication-bias estimates, and optional conditional estimates.
known_v_backend records the known-sampling-covariance backend.
inclusion_random contains random-effect inclusion tables;
selection_model, selection_sampling, and
selection_sensitivity_diagnostics retain selection-target metadata
and optional sensitivity diagnostics. The printed
form displays the non-empty tables. Objects created with only_data = TRUE
return the resolved data, and objects created with only_priors = TRUE
return a summary.brma object that prints only the model title, as at 4.0.0,
and holds the resolved prior list in its priors element, because neither
carries a posterior.
Intercept-only multivariate location
models label their sole coefficient mu, consistently with ordinary
meta-analysis models. Once location moderators are present, the coefficient
is labeled intercept; an intercept fixed at zero is omitted.
Multivariate models without mods or scale display location and random
estimates together in one Estimates table. With predictors, random estimates
join Common Estimates, location coefficients appear in Meta-Regression
(or Location when scale is present), and scale coefficients appear in
Scale. Without location moderators, mu remains in the common table.
Conditional estimates follow the same layout. Data-frame conversion follows
these displayed sections; the returned list retains its separate raw tables.
Scale tables label the baseline SD as exp(intercept); its estimates are
already exponentiated. Other scale coefficients remain on the log-SD scale.
Multivariate scale rows identify their targeted random-effect heterogeneity
(tau), total heterogeneity (tau_total), or mean-variance allocation scale
(tau_common). The baseline
corresponds to zero non-intercept design columns under the fitted contrasts
and the requested predictor standardization.
The random table reports the quantities aligned with prior specification;
use summary_heterogeneity() for aggregate variances and the complete family
of deterministic allocation transforms.
This reporting rule is the same for brma.mv() and BMA.mv(): component
tau values derived from a variance allocation appear only in
summary_heterogeneity().
Optional selection-conditioning comparisons are inspected explicitly with
selection_sensitivity_diagnostics(). Printing a model or its summary does
not emit sensitivity notifications.
Fitted objects store their default summary, so print() and summary()
without further arguments return it without recomputation; update() and
add_selection_sensitivity_diagnostics() store it again.
Examples
if (FALSE) { # \dontrun{
if (requireNamespace("metadat", quietly = TRUE)) {
data(dat.lehmann2018, package = "metadat")
fit <- bPET(
yi = yi,
vi = vi,
data = dat.lehmann2018,
measure = "SMD",
seed = 1,
silent = TRUE
)
summary(fit)
}
} # }