Description:

This Master Prompt would focus on understanding why effect sizes vary across studies rather than treating heterogeneity simply as a statistical nuisance. It would guide researchers through Cochran's Q, I², τ² and prediction intervals, followed by theoretically justified exploration of clinical, methodological, demographic, geographic or contextual sources of heterogeneity.

It would extend into subgroup analysis and univariate/multivariable meta-regression, including moderator selection, interaction interpretation, bubble plots and cautious interpretation when few studies are available. The uploaded collection contains examples with I² approaching 90–100%, subgroup analyses across study designs and populations, and meta-regression demonstrating how study-level moderators can materially alter interpretation.