Brenton M. Wiernik

I say yes. I would argue that all of these are about checking model adequacy for the factor/cluster models

I would agree--I would prefer to have all of the movement of functions happen in one cycle.

This is a method for testing contrasts of parameters using a statistic that is a Bayesian equivalent of the Wald t statistic T = Estimate / SE(Estimate). Table 2 shows...

When you say scale, do you mean the variables have the same variance/are standardized, or do you mean that the mediator and response have the same distributional family? (Not sure...

Oh, okay. Yes, that message doesn't make sense to me. In the case of a binary mediator, that would make interpreting the unstandardized indirect effect the easiest. See here for...

I see that this is apparently a point of debate in the literature, but standardization appears to do basically nothing here: https://link.springer.com/article/10.1186/s12874-018-0654-z

This was the result the first time I ran it: ```r > mM mY mediation::mediate(mM, mY, treat = "hp", mediator = "am") |> + summary() Causal Mediation Analysis Quasi-Bayesian Confidence...

(In general, I'm pretty down on mediation analyses because the necessary identification assumptions for the effect sizes [that both the treatment and all the mediators are functionally randomly assigned] are...

Oh, okay, the `mediate()` function is not appropriate for continuous treatment variables. It's approach is a "possible outcomes framework"---to simulate values from the fitted models for `treatment = 0` and...

Because logistic regression isn't invariant across score ranges. The possible outcomes framework depends on the X and M contrasts being compared.