icc: Intraclass Correlation Coefficients

In donaldRwilliams/blsmeta: Bayesian Location-Scale Meta-Analysis

Description

Compute ICCs for three-level meta-analysis Note that the ICC formulation for three-level models is described in \insertCitecheung2014modeling;textualblsmeta. It is defined as the heterogeneity at each level divided by the total heterogeneity, which will approach I2 when the sampling variances approach zero (large n).

Usage

ICC(
  object,
  newdata_scale2 = NULL,
  newdata_scale3 = NULL,
  cred = 0.95,
  summary = TRUE,
  percent = TRUE,
  digits = 3
)

Arguments

object	An object of class blsmeta.
newdata_scale2	An optional data.frame for which to compute predictions for the level 2 variance component. Defaults to NULL, which then uses the original data used in blsmeta.
newdata_scale3	An optional data.frame for which to compute predictions for the level 3 variance component. Defaults to NULL, which then uses the original data used in blsmeta
cred	numeric. credible interval (defaults to 0.95).
summary	logical. Should the posterior samples be summarized (defaults to TRUE)?
percent	logical. Should the results be percentages, as in metafor (defaults to TRUE)?
digits	numeric. The desired number of digits for the summarized estimates (defaults to 3).

Details

In essence, with a scale model, this results in the ICCs being a function of moderators. Further, rather than one ICC, there is an ICC for each of the k studies. For more information about varying ICCs, we refer interested users to \insertCitewilliams2019putting;textualblsmeta and \insertCitewilliams2020fineblsmeta.

Value

A list of two data frames of predicted values.

Examples

library(psymetadata)

fit <- blsmeta(yi = yi, vi = vi, 
               es_id = es_id,
               study_id = study_id,
               data = gnambs2020)


iccs <- ICC(object = fit, 
            summary = TRUE)

donaldRwilliams/blsmeta documentation built on Dec. 20, 2021, 12:12 a.m.