Corrected Table 1 Results for different indicators to assess the performance of the three meta-analysis bias-adjustment methods

From: Author Correction: Comparing meta-analyses and preregistered multiple-laboratory replication projects

Method

False positive rate, 0.5% level (5% level)

False negative rate, 0.5% level (5% level)

Mean meta-replication difference; random effects (z-statistic; P value)

Mean meta-replication difference; Random Effects 99.5% CI (95% CI)

Mean meta-replication difference; unweighted

Over-estimation factor

Root mean squared error

Mean MDE 0.5% level (5% level)

Random Effects

100% (100%)

0% (0%)

0.261 (5.46; <0.001)

0.13, 0.40 (0.17, 0.35)

0.26

2.59

0.31

0.21 (0.16)

PET-PEESE

14.3% (16.7%)

42.9% (50%)

0.067 (0.67; 0.503)

-0.21, 0.35 (-0.13, 0.26)

0.05

1.31

0.26

0.45 (0.35)

3PSM

71.4% (83.3%)

28.6% (12.5%)

0.210 (4.72; <0.001))

0.09, 0.34 (0.12, 0.30)

0.20

2.24

0.25

0.31 (0.24)

Trim-and-fill

100% (100%)

0% (0%)

0.236 (5.59; <0.001)

(0.12, 0.36) (0.15, 0.32)

0.23

2.43

0.28

0.21 (0.16)

  1. In estimating the indicators, the meta-analytic results for the bias-adjustment methods are compared to the results for the replication studies (we also include results for the Random Effects model as a benchmark of the value of the indicators for uncorrected meta-analysis). Results for several indicators are shown for both the 0.5% and the 5% significance levels; see the Methods section for the definition of the indicators.
Back to article page