In this article (Leifeld and Cranmer Reference Leifeld and Cranmer2019a), three of the thirteen model terms in the TERGM in Section 4 are contaminated by observations originating from the time step that is held out for out-of-sample assessment of model fit. This circularity invalidates the model comparison in the empirical example that was used for illustration.
The authors would like to thank Block et al. (Reference Block, Hollway, Stadtfeld, Koskinen and Snijders2022) for pointing to the error and refer readers to their response to Block et al. (Leifeld and Cranmer Reference Leifeld and Cranmer2022). This exchange is briefly discussed in the Editors’ Note by Wasserman and Brandes (Reference Wasserman and Brandes2022).
The authors would also like to note to readers that the replication data for the original study is available in the Dataverse repository (Leifeld and Cranmer Reference Leifeld and Cranmer2019b).
