A note on variance estimation in random effects meta-regression.

Area of Science:

• Biostatistics
• Epidemiology
• Statistical Inference

Background:

• Meta-regression analysis commonly uses estimated weights, potentially leading to incorrect covariance matrices.
• Accurate variance estimation is essential for reliable parameter estimates in random effects meta-regression.

Purpose of the Study:

• To investigate a robust variance estimation approach for random effects meta-regression.
• To compare the performance of robust variance estimation with other methods using simulation and real-world data.

Main Methods:

• A robust variance estimation approach treating the assumed covariance matrix as a working matrix.
• Illustration using meta-analysis data from vaccine clinical trials.
• Simulation study assessing bias and coverage probability of different variance estimators.

Main Results:

• The robust variance estimation approach was investigated for random effects meta-regression.
• Comparison with two other methods, including the Knapp and Hartung (2003) estimator.
• The Knapp and Hartung (2003) estimator demonstrated superior performance in simulations.

Conclusions:

• Despite the theoretical appeal of robust estimators, the Knapp and Hartung (2003) method provides the best protection against errors in estimated weights in random effects meta-regression.
• This improved estimator is recommended for more reliable statistical inference in meta-regression analyses.