Brian Harvey Willis

This paper compares closed-form and numerical methods for estimating initial values in a bivariate random effects model for meta-analysis of test accuracy studies, finding closed-form methods reduce computation time by 80%.