Abstract
I strongly support Carlin and Moreno-Betancur's assertion that regression modeling (and in fact, any modeling) should be driven by the type of research question: descriptive, predictive, or causal. I share six points to highlight and clarify further confusions and suggest additional tricks to identify the right type. These include a focus on actions, individual vs. collective, additional causal typologies, the significance of variable names, software-related issues, and the role of data.