Table 1 The role of the standard error in 7 benchmark estimators

Estimator	Weight	Regressor	Identification
Simple average
FE/WLS	✓
PET-PEESE	✓	✓
EK	✓	✓	✓
WAAP	✓		✓
Andrews & Kasy	✓		✓
p-uniform^*	✓		✓

Notes: Simple average = unweighted mean. FE/WLS = fixed effect/weighted least squares: mean weighted by inverse variance³². PET-PEESE = precision-effect test and precision-effect estimate with standard errors: selection is a quadratic function of SE when true effect ≠ 0⁷. EK = endogenous kink: selection is a linear function of SE for imprecise estimates, no selection for precise ones⁹. WAAP = weighted average of adequately powered estimates: only estimates with at least 80% power included⁸. Andrews & Kasy = a selection model in the tradition of Hedges¹⁴. p-uniform^* = a simplified selection model based on the principle that p-values should be uniformly distributed at the true effect size¹⁵.

Quick links

Search