How to model a multivariate multilevel linear model

Questions
5

Thanks chartl, I’am pretty new to the pymc and theano, how to using this model to my data to get the
trace? now I write the mu like this:

mu = pm.Deterministic('mu', alpha[site_idx] + beta_mat[0][site_idx]*all_site_train['辐照度'] + beta_mat[1][site_idx]*all_site_train['风速'] +
                     beta_mat[2][site_idx]*all_site_train['风向'] + beta_mat[3][site_idx]*all_site_train['温度'] +
                     beta_mat[4][site_idx]*all_site_train['湿度'] + beta_mat[5][site_idx]*all_site_train['压强'])
y_pred = pm.Normal('y_pred', mu=mu, sd=sigma, observed=y_train_scaled)

still a naive way…