metarep {metarep}

Replicability-analysis of a meta-analysis

Description

Add results of replicability-analysis to a meta-analysis, whether common- or random-effects.

Usage

metarep(
  x,
  u = 2,
  t = 0.05,
  alternative = "two-sided",
  report.u.max = FALSE,
  confidence = 0.95,
  common.effect = FALSE
)

Arguments

x

object of class 'meta'

u

replicability requirement. u must be an intiger between 2 and n (nmber of studies in the meta-analysis).

t

truncation threshold for truncated-Pearsons' test ('t=0.05' by default). t is ignored if 'common.effect = TRUE'.

alternative

use 'less', 'greater' or 'two-sided'

report.u.max

use TREU to report the lower bounds on number of studies with replicated effect.

confidence

Confidence level used in the computaion of the lower bound(s) u_{max}^L and\or u_{max}^R.

common.effect

Use common.effect = FALSE (default) for replicability-analysis combining with no assumptions (Pearson or truncated-Pearson test). Replicability-analysis based on the test-statistic of common-effects model can be applied using common.effect = TRUE.

Value

An object of class list containing meta-analysis and replicability analysis results, as follows:

worst.case.studies

A charachter vector of the names of n-u+1 studies at which the the r(u)-value is computed.

r.value

r(u)-value for the specied u.

side

The direction of the effect with the lower one-sided r(u)-value

u_L, u_R

Lower bounds of the number of studies with decreased or increased effect, respectively. Both bounds are reported simultinualsly only when performing replicability analysis for two-sided alternative with no assumptions

Examples

 n.i.1 <- c( 20, 208, 24, 190, 58, 36, 51)
a.i <- c( 2,79,0,98,15,34,9) 
n.i.2 <- c( 20, 119, 22, 185, 29, 51, 47)
c.i <- c(9,106,14,98,12,49,9) 
m1 <- metabin( event.e = a.i,n.e = n.i.1,event.c = c.i,n.c = n.i.2,
               studlab = paste0('Study ' , 1:7) , sm = 'OR' ,
               common = FALSE, random = TRUE )
mr1 <- metarep(  m1 , u = 2, common.effect = FALSE , t = 0.05 , 
               alternative = 'two-sided', report.u.max = TRUE)
forest(mr1, layout='revman5',digits.pval = 4 , test.overall = TRUE )