Open wjakethompson opened 1 year ago
For LCDM, we could specify all possible parameters, and then use an expanded Q-matrix (i.e., with interaction terms calculated) to zero out relevant parameters (e.g., l1_212 = raw_l1_212 * Q[2, {column}].
l1_212 = raw_l1_212 * Q[2, {column}]
Would need to also consider how to pass user-specified priors. rstanarm uses pre-compiled models and allows user-specified priors, so maybe look at implementation here: https://github.com/stan-dev/rstanarm/tree/master/src/stan_files
For LCDM, we could specify all possible parameters, and then use an expanded Q-matrix (i.e., with interaction terms calculated) to zero out relevant parameters (e.g.,
l1_212 = raw_l1_212 * Q[2, {column}].Would need to also consider how to pass user-specified priors. rstanarm uses pre-compiled models and allows user-specified priors, so maybe look at implementation here: https://github.com/stan-dev/rstanarm/tree/master/src/stan_files