Introduction to the pimeta package

Kengo Nagashima

2018-09-15

The pimeta package is easy. Load your data and then pass it the pima function!

require("pimeta")
data(sbp, package = "pimeta")

# a parametric bootstrap prediction interval
set.seed(20161102)
pima(y  = sbp$y,      # effect size estimates
     se = sbp$sigmak, # within studies standard errors
     B  = 50000       # some options (e.g., number of bootstrap samples)
)

> # a parametric bootstrap prediction interval
> set.seed(20161102)
> pima(y  = sbp$y,      # effect size estimates
+      se = sbp$sigmak, # within studies standard errors
+      B  = 50000       # some options (e.g., number of bootstrap samples)
+ )

Prediction Interval for Random-Effects Meta-Analysis

A parametric bootstrap prediction interval
 Heterogeneity variance: DerSimonian-Laird
 SE for average treatment effect: Hartung

Average treatment effect [95%PI]:
 -0.3341 [-0.8769, 0.2248]

Average treatment effect [95%CI]:
 -0.3341 [-0.5660, -0.0976]

Heterogeneity variance (tau^2):
 0.0282

Several type of methods are supported.

# Higgins-Thompson-Spiegelhalter prediction interval
pima(sbp$y, sbp$sigmak, method = "HTS")

# Partlett-Riley prediction interval (Hartung and Knapp's SE)
pima(sbp$y, sbp$sigmak, method = "HK")

# Partlett-Riley prediction interval (Sidik and Jonkman's SE)
pima(sbp$y, sbp$sigmak, method = "SJ")

> # Higgins-Thompson-Spiegelhalter prediction interval
> pima(sbp$y, sbp$sigmak, method = "HTS")

Prediction Interval for Random-Effects Meta-Analysis

Higgins-Thompson-Spiegelhalter prediction interval
 Heterogeneity variance: DerSimonian-Laird
 SE for average treatment effect: standard

Average treatment effect [95%PI]:
 -0.3341 [-0.7598, 0.0917]

Average treatment effect [95%CI]:
 -0.3341 [-0.5068, -0.1613]

Heterogeneity variance (tau^2):
 0.0282

> 
> # Partlett-Riley prediction interval (Hartung and Knapp's SE)
> pima(sbp$y, sbp$sigmak, method = "HK")

Prediction Interval for Random-Effects Meta-Analysis

Partlett-Riley prediction interval
 Heterogeneity variance: REML
 SE for average treatment effect: Hartung-Knapp

Average treatment effect [95%PI]:
 -0.3287 [-0.9887, 0.3312]

Average treatment effect [95%CI]:
 -0.3287 [-0.5761, -0.0814]

Heterogeneity variance (tau^2):
 0.0700

> 
> # Partlett-Riley prediction interval (Sidik and Jonkman's SE)
> pima(sbp$y, sbp$sigmak, method = "SJ")

Prediction Interval for Random-Effects Meta-Analysis

Partlett-Riley prediction interval
 Heterogeneity variance: REML
 SE for average treatment effect: Sidik-Jonkman

Average treatment effect [95%PI]:
 -0.3287 [-0.9835, 0.3261]

Average treatment effect [95%CI]:
 -0.3287 [-0.5625, -0.0950]

Heterogeneity variance (tau^2):
 0.0700