michaelwozniak / reproducible_research_project

Reproducible Research project of the working paper Institutional Underpinnings and Democracy Backsliding in the Perspective of COVID-19

Requirements/Instalation

How to install python dependencies:

pip install -r py_requirements.txt

How to install R dependencies:

install.packages(scan(file = "r_requirements.txt", what = character()))

Project tree

|-- README.md
|-- codes
|   |-- _init_.py
|   |-- experiments
|   |   |-- _init_.py
|   |   |-- main_research
|   |   |   |-- 1_data_merging_and_cleaning.ipynb
|   |   |   |-- 2_ols_regression.ipynb
|   |   |   |-- 3_catboost_regression.ipynb
|   |   |   |-- 4_random_forest_regression.ipynb
|   |   |   |-- _init_.py
|   |   |   `-- catboost_info
|   |   `-- robustness_check
|   |       |-- 1_data_merging_and_cleaning.ipynb
|   |       |-- 2_ols_regressions.ipynb
|   |       |-- 3_catboost_regression.ipynb
|   |       |-- 4_random_forest_regression.ipynb
|   |       |-- _init_.py
|   |       `-- catboost_info
|   `-- modules
|       |-- _init_.py
|       |-- py_modules
|       |   |-- _init_.py
|       |   |-- ale.py
|       |   |-- feature_engineering.py
|       |   |-- ml_models_training.py
|       |   `-- xai_wrapper.py
|       `-- r_modules
|           |-- diagnostic_tests.R
|           `-- feature_engineering.R
|-- data
|   |-- main_research
|   |   |-- processed_data
|   |   `-- raw_data
|   `-- robustness_check
|       |-- processed_data
|       `-- raw_data
|-- py_requirements.txt
|-- r_requirements.txt
`-- results
    |-- main_research
    |   |-- catboost
    |   |-- ols_regression
    |   `-- random_forest
    `-- robustness_check
        |-- catboost
        |-- ols_regression
        |   |-- main_tables
        |   |-- robustness_collinearity
        |   |-- robustness_missings
        |   `-- robustness_time_periods
        `-- random_forest

Project description

We created a project based on the article "Institutional Underpinnings and Democracy Backsliding in the Perspective of COVID-19" and the poster "A study of Democracy Backsliding in the Perspective of COVID-19 using econometrics and machine learning approaches" by PhD Jacek Lewkowicz, Michał Woźniak and Michał Wrzesiński. Originally we investigated the drivers of the impact of the political decisions, as a response to the COVID-19 pandemic. We used a novel global dataset covering the period of the first wave of the pandemic (March-June 2020), which included the variables that we used as dependent variables. In these studies, we applied various econometric and machine learning tools.

The main goals of the reproducible project are:

• reproduction of the initial results
• improvement of the study by performing robustness checks
• structuring the project
• clean and readable code for reproducibility
• combination of the research methodology developed as a part of a scientific article and a poster from conference MLinPL

Project structure description

We divide our project into three main folders.

1. In code folder you can find experiments and modules. Folder modules contains python and R classes and functions. Folder experiments is divided into two subfolders: main_research and robustness_check. Folder main_research contains files that were created during original research. In turn, folder robustness_check contains files, that are reproduced by us with new dependent variables (updated by the authors for a longer time range).
2. In data folder you can find two subfolders for main research and robustness check. In main_research, folder raw_data contains initial files that were used to create database. In turn, processed_data contains final_dataset.csv that is our main dataset for initial study. Analogously, robustness_check folder has two subfolders for raw data and processed data. raw_data contains final dataset from main research, new dataset for dependent variables and dataset for covid data. Folder processed data contains four processed files: final_dataset_all for new time range (March-December 2020) and minor datasets for three subperiods (March-June 2020, July-September 2020 and October-December 2020). In both raw_data folders for main research and robustness check you can find variables_description.md files with short description of databases/variables.
3. In result folder, again you can find two subfolders for main research and robustness check. Both of them contains results for OLS regression, Random Forest and CatBoost. In the case of OLS regressions for main_research there are two tables for main regressions. For robustness_check we have division into four folders: main_tables (reproduced tables from main research with new dependent variables), robustness_time_periods, robustness_collinearity and robustness_missings. All above mentioned folders contains tables with regression results for various specifications. In the case of Random Forest and CatBoost, both main_research and robustness_check folders contain analogous files: results for Permutation Importance, SHAP summary plots, Partial Dependence Plots (1 dimensional and 2 dimensional) and Accumulated Local Effects plots (1 dimensional and 2 dimensional).