Circular specifications and “predicting” with information from the future: Errors in the empirical SAOM–TERGM comparison of Leifeld & Cranmer

Abstract We review the empirical comparison of Stochastic Actor-oriented Models (SAOMs) and Temporal Exponential Random Graph (TERGMs) by Leifeld & Cranmer in this journal [Network Science 7(1):20–51, 2019]. When specifying their TERGM, they use exogenous nodal attributes calculated from outcome networks’ observed degrees instead endogenous ERGM equivalents structural effects as used SAOM. This turns modeled endogeneity into circularity obtained results are tautological. In consequence, out-of-sample predictions using TERGMs based on information thereby predict future observations future. Thus, analysis rests erroneous model specifications that invalidate article’s conclusions. Finally, beyond these specific points, we argue evaluation metric—tie-level predictive accuracy—is unsuited for task comparing performance.