I recently came across the idea of narratively generative modeling and have been trying to understand it more deeply, especially the emphasis on specifying a full data-generating process and interrogating it through prior and posterior predictive checks. Michael Betancourt’s Reading Times case study (see below) is in my point of view a really nice example of this workflow in practice, showing how model assumptions translate into implied data and how that shapes inference. How many of you here use this kind of modeling in your own line of work and in what contexts has it proven most useful?
3 Likes
Nice article. Has some relationship to U statistics and to Handbook of Generalized Pairwise Comparisons | Methods for Patient-Cen