PublicationBiasBenchmark (version 0.1.3)

method.AK: AK Method

Description

Implements the Andrews & Kasy (AK) method for publication bias correction in meta-analysis. The AK method categorizes estimated effects into groups with different probabilities of being published. AK1 uses symmetric selection grouping estimates into significant (|t| >= 1.96) and insignificant (|t| < 1.96) estimates. AK2 uses asymmetric selection with four groups based on both significance and sign: highly significant positive/negative effects and marginally significant positive/negative effects, each with different publication probabilities. See andrews2019identification;textualPublicationBiasBenchmark for details.

Usage

# S3 method for AK
method(method_name, data, settings)

Value

Data frame with AK results

Arguments

method_name

Method name (automatically passed)

data

Data frame with yi (effect sizes), sei (standard errors), and study_id (for clustering wherever available)

settings

List of method settings (see Details)

Details

The following settings are implemented

"default"

Uses AK1 estimator (symmetric selection)

"AK1"

Symmetric selection model grouping estimates into significant (|t| >= 1.96) and insignificant (|t| < 1.96) categories with relative publication probabilities of 1 and p1 respectively.

"AK2"

Asymmetric selection model with four groups based on t-statistics: (a) t >= 1.96, (b) t < -1.96, (c) -1.96 <= t < 0, and (d) 0 <= t < 1.96, with relative publication probabilities of 1, p1, p2, and p3 respectively.

References

Examples

# Generate some example data
data <- data.frame(
  yi = c(0.2, 0.3, 0.1, 0.4, 0.25),
  sei = c(0.1, 0.15, 0.08, 0.12, 0.09)
)

# Apply AK method
result <- run_method("AK", data, "default")
print(result)