Random-Effects Meta-Analysis of Mash Pairwise Contrasts, per Condition

For each condition (context), gathers all pairwise contrast effect sizes and standard errors involving that condition, then runs a random-effects meta-analysis (via metafor::rma; pecotmr does not implement its own meta-analysis). Intended to be run on the output of fitMashContrast / mashPosteriorContrast.

Usage

metaAnalysisPerCondition(
  effectSizes,
  seValues,
  seCutoff = 0,
  metaMethod = "DL"
)

Arguments

effectSizes: Numeric matrix (features x contrasts) of contrast effect sizes. Column names must follow the pattern mean_contrast_<conditionA>_vs_<conditionB>.
seValues: Numeric matrix (features x contrasts) of contrast standard errors. Must have the same dimensions and column names as effectSizes.
seCutoff: Numeric; minimum SE below which a contrast is excluded from the meta-analysis for a given feature (default 0).
metaMethod: Between-study variance estimator for the random-effects meta-analysis, forwarded to metafor::rma(method = ). Default "DL" (DerSimonian-Laird); other options include "REML", "ML", "EB".

Value

A tibble with columns:

condition: The condition (context) name.
contrast: Pairwise contrast name (without prefix), e.g. conditionA_vs_conditionB.
meta_pvalue: P-value from the random-effects meta-analysis.
meta_effect: Pooled absolute effect size estimate.
meta_se: Standard error of the pooled estimate.
tau2: Between-study variance estimate.
I2: Heterogeneity measure (proportion of variance due to between-study variance), in [0, 1].