Capturing heterogeneity and PLS-SEM prediction ability: Alliance governance and innovation

Whether the uses of PLS-SEM latent interaction effect (PLS-LIE), PLS prediction-oriented segmentation (PLS-POS), or PLS-PATHMOX approaches improve prediction ability remains unclear. The present study draws on holdout sample estimations to assess prediction ability for the three approaches. The illustrative empirical model focuses on examining the differences in innovation outcomes that coopetive alliances produce, in which three governance factors may affect the structural model parameter estimates reflecting these differences. The findings suggest improvements in prediction ability with the use of PLS-LIE and PLS-POS but not for PLS-PATHMOX.