This repo features a classical regression analyses with several predictors, but uses a Bayesian approach to answer the underlying research question. Using Bayesian analyses has several advantages. Among these are the independence of classical p-values and the incorporation fo own metrics as well as the use of prior knowledge that affects the prior and posterior distribution in a certain direction before the analyses. This feature is mainly important when considering that research should be a cumulative process, where prior research is rather incorporated than neglected!
To show the transparent process of a full Bayesian analyses, no pre-built packages are used (except for Bayes factor). In this way, everyone can have a look at the code to further understand the underlying constructs or re-use it for their own analyses.
In this project NBA salary and Stats data is used to investigate a common question in all team sports.
Research Question: Are there differences in the valuation of offensive and defensive game statistics in professional basketball and which game statistics are valued the most?
The primary data is available on Kaggle. The original data and further data used in this analyses is scraped from this website.
This repo contains several files for the analyses of this research question. Mainly the code does not change between the documents, but for easier readibility several documents are created and analyses are run. Moreover, knitted documents are created, that easen up the overview. The main analyses without the manuscript are available here and the main manuscript generated by the code is available here.
| Files/Folders | Description |
|---|---|
| 1. Full Code | R-Markdown-script with full functions and output |
| 2. Analyses | Main Analyses: Rendered to a pdf version for easy code overview |
| 3. Manuscript | R-Markdown-script to obtain Manuscript with minimal code to answer RQ |
| /Data | Folder containing scraped data. One for main analyses and one for historical data |
| /Functions | Folder containing helper functions. Needed for all analyses! |
| /Manuscript | Folder containing pdf versions of Rmd files |
For any help with the files in this archive, please contact Christoph Völtzke (voeltzke.christoph@gmail.com).
The repository is licensed under the apache-2.0 license.