Non-identifiability for Gaussian process with random intercepts (https://discourse.mc-stan.org/t/non-identifiability-for-gaussian-process-with-random-intercepts/29693)

.gitignore

.ipynb_checkpoints
*.hpp
model-*
!model-*.stan
????-??-??-*.ipynb

data.stan

// Common data definitions for all models.

data {
    int num_obs, num_units, exp_transform;
    array [num_obs] real x;
    array [num_units] vector[num_obs] ys;
    real<lower=0> sigma, length_scale, kappa;
}

transformed data {
    cov_matrix[num_obs] cov = add_diag(gp_exp_quad_cov(x, sigma, length_scale), 1e-9);
    matrix[num_obs, num_obs] chol = cholesky_decompose(cov);
}

Makefile

requirements.txt :requirements.in
	pip-compile -v

inference.ipynb

model-cheat-non-centered.stan

model-non-centered-hard-intercepts.stan

model-non-centered-soft-intercepts.stan

model-non-centered-soft-zero-gp.stan

model-non-centered.stan

observation_model.stan

requirements.in

requirements.txt