hierarchical-matlab

Matlab implementation of Hierarchical Bayesian Inference

Steps for a complete Hierarchical Bayesian sampling with propagation of uncertainty in the predictions

step 0: synthetic data

• go to engines/ and run make_data.m

step 0.1: optimize (optional)

• go to /optimize/
• run optimize_theta_ind.m
• compare the inferred variables in CMA with the nominal in data.data.theta and data.data.std_data

step 1: posterior of individual theta_i

• go to /sample/ and run Sample_theta_ind.m (it will take some time)
• go to ../postprocess/ and run "load ../../data/IND_theta_002.mat; plotmatrix_hist(out_master.theta);" to see histogram of the parameters for the 2nd data set

step 2: sample hyperparameters

• go to /sample/ and run Sample_psi.m
• go to /postprocess/ and run "load ../../data/HB_unif_psi.mat; plotmatrix_hist(out_master.theta);" to see a histogram of psi

step 3a: post-process

• go to /postprocess/ and run prepare_data_post_theta.m

step 3b: posterior of theta_i

• go to /sample/ and run Sample_theta_post.m

step 4: propagate uncertainty

• go to /propagate/
• run run_model.m to run the model with theta samples
• run uq_bounds to plot the uncertainty in the prediction