Quick question about # of parameters for assessing p_loo

Dear all,

Just want to check my understanding of how one counts the # of parameters for assessing p_loo. Assume we have hierarchical model with 20 students nested in 50 schools. Assume also that the outcome is regressed on a student characteristic (e.g. SES) and that there is an intercept and slope for each of the 50 schools modeled as a function of some school level variable (e.g. public v. private school). In addition to the school level (so-called) “fixed effects”, I presume that I also count the 50 school level intercept values and 50 slope values, as well as within residual variances. By my count I would have 50 intercepts, 50 slopes, 50 residual variances, a school level intercept, a school level slope, and a school level residual term (153). Is my thinking correct?

Thank you


Looks correct (although I have always difficulties following the fixed/random effect language instead of hierarchical model language with common parameters, population prior, and group specific parameters).