Models for estimating the number of unpublished studies

Area of Science:

• Biostatistics
• Scientific Research Methodology

Background:

• Publication bias, particularly against non-significant results, can compromise the validity of meta-analytic conclusions.
• The existence of unreported studies poses a significant challenge to the reliability of literature summaries.

Purpose of the Study:

• To propose statistical models for estimating the number of unpublished studies.
• To provide meta-analysts with tools to assess the impact of publication bias on study conclusions.

Main Methods:

• Development of two general statistical models.
• Utilizing reported p-values from published studies to estimate the quantity of unpublished studies (N).
• Evaluating the plausibility of estimated N and associated confidence bounds.

Main Results:

• The proposed models offer a quantitative approach to estimating unpublished studies.
• Meta-analysts can assess the potential impact of publication bias using these estimation techniques.

Conclusions:

• These models enable meta-analysts to evaluate the problem of unpublished studies from multiple viewpoints.
• Increased understanding and confidence in meta-analytic conclusions can be achieved through the application of these methods.