Group info

General
1

this link has group predictors :

https://widdowquinn.github.io/Teaching-Stan-Hierarchical-Modelling/11-group_level_predictors.html

it shows the group predictor “u” is the same length as the the number of observations N so all the rows from a county will have the same value. Is that the correct way to add a group predictor or should the group predictor be the same length as the group?

seems the group predictor is at the group level not the observation level so why have it at observation level?

And in this link 19.1.3 it states “we need only plunk the group-level walk_score predictor directly into the stan_glmer() formula. Since all listings in the same neighborhood share the same walk_score , it is automatically recognized as a group-level predictor.”

Will stan glmet have the group predictor the length of data or length of the group levels?

2

If coding a model in Stan directly you could do it either way but because rstanarm only accepts one data frame all the variables need to be the same length. If they only vary by group then there will just be repeated values in order to make the length correct.