Hybrid Clinical Trial Analysis

Combining Bayesian Inference and Unsupervised Learning

Simulation Controls

Adjust the parameters below and click "Run Analysis" to see how they impact the results and visualizations.

Gender Filter

Introduction

This dashboard combines two powerful analytical methods on the same simulated clinical trial data. **Bayesian Inference** (MCMC) provides a precise, probabilistic framework for answering specific questions about a drug's effectiveness, while a **Variational Autoencoder** (VAE) offers an unsupervised, visual exploration of the data to uncover hidden patterns and relationships. By viewing both perspectives, we gain a comprehensive understanding of the trial's outcomes. In this simulation, the goal of the treatment is to **lower** blood pressure, so a more negative effect indicates better performance.

Confirmatory Analysis (Bayesian)

This section presents the results of a hierarchical Bayesian model, similar to what you'd get from a PyMC analysis. The table below provides the posterior summary for the effects of each treatment and demographic factor, including the mean and a credible interval. The probabilistic statements below offer direct, intuitive answers to key questions about the drug's performance.

Parameter Mean SD HDI 3% HDI 97%

Exploratory Analysis (VAE)

This section shows the data organized in a 2-dimensional latent space, as learned by a VAE. The plot below visually confirms the relationships between treatment groups and demographic factors. Patients with similar profiles are clustered together, providing a compelling visual intuition of the drug's effects without relying on formal statistical tests. The tooltips show individual patient data, allowing you to explore the clusters more closely.