Between-Study Heterogeneity
Slides
Keywords: Cochran’s Q, I-Squared, Prediction Intervals, Outliers, Baujat Plot
Code
# This data set contains effects of studies examining
# the effects of suicide prevention interventions
# Copy and paste the entire code below into your Console.
# Then hit Enter.
data <- structure(
list(
study = c("Berry et al.", "DeVries et al.",
"Fleming et al.", "Hunt & Burke",
"McCarthy et al.", "Meijer et al.",
"Rivera et al.", "Watkins et al.",
"Zaytsev et al."),
es = c(-0.143, -0.608, -0.111, -0.127, -0.392,
-0.268, 0.012, -0.245, -0.126),
se = c(0.147, 0.17, 0.258, 0.176, 0.202,
0.135, 0.183, 0.224, 0.194)),
row.names = c(NA, -9L),
class = "data.frame")
# Random-Effects Model
# If necessary, install the meta package
install.packages("meta")
# Load the meta package
library(meta)
# Using our 'data' dataset, run metagen:
rem <- metagen(es, se, study, data)
summary(rem)
forest(rem)
# Prediction Intervals
# Set prediction to TRUE
rem <- metagen(es, se, study, data, prediction = TRUE)
summary(rem)
forest(rem)
# Outlier Removal
# Exclude study by de Vries (no. 2)
rem <- metagen(es, se, study, data[-2,], prediction = TRUE)
summary(rem)
forest(rem)