Bayesian Logistic Analysis

This project applies Bayesian logistic regression to a social network purchase dataset. The analysis is implemented in R and leverages Bayesian inference to assess model parameters, quantify uncertainty, and generate posterior predictions.

Data Source

• Dataset: Social Network Purchase Data
• Author: ANANYA NAYAN, 2017

Project Overview

In this analysis, you will find:

• Data Wrangling: Importing and cleaning the dataset to prepare it for modeling.
• Model Building: Fitting a Bayesian logistic regression model to predict the binary outcome (Purchased or Not Purchased) based on predictors such as Estimated Salary and Age.
• Posterior Inference: Evaluating the posterior distributions of the model parameters, computing credible intervals, and checking convergence diagnostics (e.g., R-hat values and effective sample sizes).
• Visualization: Generating plots to illustrate the posterior distributions and model predictions.

Imported Libraries

This R notebook loads several key packages to support data manipulation, modeling, and visualization. For example, the notebook includes the following libraries:

• dplyr – for data manipulation
• ggplot2 – for creating visualizations
• rstanarm – for fitting Bayesian generalized linear models
• bayesplot – for plotting posterior distributions and diagnostics
• bayesrules – for applying Bayesian methods
• tidyverse – for a collection of data science tools
• tidybayes – for tidying Bayesian model outputs
• broom.mixed – for summarizing model results in a tidy format

These packages are loaded at the beginning of the notebook, as shown in the code chunk (see :contentReference[oaicite:0]{index=0}).

How to Run the Analysis

1. Download or Clone the Repository:
   Get a copy of the project repository containing the R notebook.

2. Open the Notebook:

   • If you prefer a static view, open the Bayesian Logistic Analysis.nb.html file in your web browser.
   • For an interactive experience, open the corresponding R Markdown (.Rmd) file in RStudio.

3. Execute the Analysis:
   Follow the step-by-step instructions in the notebook to load the data, build the Bayesian model, and review the results.

Results and Interpretation

The notebook provides:

• Parameter Estimates: Posterior means, credible intervals, and diagnostic statistics (such as R-hat and effective sample size) for the model coefficients.
• Posterior Predictions: Visualizations that illustrate the predicted probabilities for purchasing behavior based on input predictors.
• Model Diagnostics: Insights into model convergence and fit.