A simple confidence interval for meta-analysis.

Area of Science:

• Biostatistics
• Medical Research Methodology
• Statistical Modeling

Background:

• Estimating confidence limits for the overall mean effect is crucial in random effects meta-analysis.
• The DerSimonian and Laird method is a common approach, but its coverage probability can be suboptimal.
• Existing statistical literature offers various methods, some requiring complex iterative computations.

Purpose of the Study:

• To introduce a simple, alternative method for constructing confidence intervals in random effects meta-analysis.
• To evaluate the performance of this new method against the established DerSimonian and Laird approach.
• To provide a computationally straightforward alternative for meta-analysis confidence interval estimation.

Main Methods:

• A novel approach for confidence interval construction based on the t-distribution was developed.
• The method was compared to the DerSimonian and Laird approach in the context of random effects models.
• The proposed method avoids iterative computations, enhancing ease of calculation.

Main Results:

• The t-distribution-based confidence interval method demonstrated improved coverage probability compared to the DerSimonian and Laird method.
• The new approach is computationally simple and does not require iterative calculations.
• This method offers a practical and statistically sound alternative for meta-analysis.

Conclusions:

• The proposed t-distribution method is a superior and simpler alternative for estimating confidence intervals in random effects meta-analysis.
• This approach enhances the reliability of meta-analysis findings by providing better coverage probability.
• The non-iterative nature of the method makes it easily accessible for researchers.