This is a chance observation due to some current work but I wondered if the 2 in the following statement isn’t an arbitrary constant?
weight = inv_logit(2 * (i - tau));
So far as I can see it governs the speed of the switch-over and should (for example) 10 be used instead the answer may be different.
Would it be a good idea to use a larger value here (e.g. 10,20) to reduce the smoothing effect whilst keeping the model continuous?