SBART Implementation on AgrImOnIA dataset

Group project for the Bayesian Statistics course in POLIMI (Politecnico di Milano) A.A. 2023/2024.

Main goal

Apply the model in Kim (2022), using the R code there for MCMC, for the analysis and prediction of air pollution data in Lombardy (AGRIMONIA data)

Get started

To run the project you need to have R installed on your machine. You can download it from here. Additionally, you need to install the R package manager renv.

Install dependencies

To install the dependencies you need to run the following command in the R console:

 renv::restore()

Run the project

You can either run the console version by executing "main.R" file or Run the more user friendly version.

Console version

 renv::run("main.R") 

Datasets

Remember to put the datasets AGC_Dataset_v_3_0_0.csv and Agrimonia_Dataset_v_3_0_0.csv in the folder data/AgrImOnIA/raw.

Tune it up

Model parameters

You can tune the model by changing the parameters in the config.R file. The parameters are:

• n_iterations: number of iterations for the MCMC algorithm.
• n_trees: number of trees in the BART model.
• warmup: number of warmup iterations for the MCMC algorithm (See Kim article for more information).

Additionally, you can change other parameters like:

• model_filename: name of the file where the results will be saved.
• date_begin: starting date to cut the dataset.
• date_end: ending date to cut the dataset.
• response_variable: name of the response variable in the dataset.
• covariates_of_interest: names of the covariates in the dataset.

Results

When the model is run, the results are saved in the output folder. The results consist of: - covariates_selection_chain: covariates selection through the iterations. - spatial_theta_chain: spatial theta through the iterations. - sigma2_chain: sigma2 through the iterations. - trees_chain: trees through the iterations. - w_selection_chain: Weight matrix selection through the iterations. - dt_history: Tree structures history. - y_predictions: Final prediction history. - y_predictions_history: Predictions history.

The information will be saved in a .RData file, the name of the file can be changed in the config.R file.