Ridge estimation in network meta-analysis

This paper explores ridge regression for treatment ranking in network meta-analysis. We show how the dependency structure of the experimental graph complicates shrinkage estimation, making classical tuning methods, such as crossvalidation, unreliable. To address this, we propose an alternative approach based on minimizing the total mean square error evaluated at pilot estimates obtained via restricted maximum likelihood estimation. Extensions to multi-arm trials and focused ridge estimation will be discussed during the conference presentation.