Regression analysis has been employed previously to express Visual Evoked Potential (VEP) VA in terms of its subjective equivalent. Bland-Altman analysis was also used in the modelling process to check for good agreement between test methods before deriving the regression equation. However, application of these methods to the full range of electrophysiological/subjective VA comparisons resulted in few making the cut for full analysis. This article considers the statistical parameters that were used in the past, and how they could be relaxed to increase inclusion in meta-analysis. The regression coefficient of VEP VA is the recommended ‘effect size’, and where a more sophisticated meta-analysis is possible, the degree of heteroscedasticity is proposed as a second variable.
