Formative Measurement Models: A Response to Bainter & Bollen (2014) and Howell (2014)

摘要

This commentary is a response to Bainter and Bollen (2014) and Howell (2014). We challenge two claims: (a) Bainter and Bollen argue that the latent variable defined by causal indicators (possessing conceptual unity) do not suffer from instability of the formative measurement model. Bainter and Bollen performed simulations to justify their claim. We applaud Bainter and Bollen's work because it seems that these Monte Carlo simulations are the first simulations published for the purpose of discussing the stability of formative measurement models; (b) in contrast, Howell considers that the causal indicators do not explain the meaning of the formative latent variable because this meaning "is entirely in the ys" (p. 142).