How does the skill handle correlated predictors?

It provides a QR decomposition implementation for Stan which decorrelates the predictors during sampling, significantly improving Markov Chain Monte Carlo (MCMC) efficiency.

Does this skill support both Stan and JAGS?

Yes, it provides side-by-side implementation patterns for both Stan and JAGS across linear, logistic, and Poisson regression models.

What is the benefit of robust regression in this skill?

The robust regression pattern uses a Student-t distribution instead of a normal distribution, which allows the model to be less sensitive to outliers in the data.

Does it provide guidance on choosing priors?

Yes, it includes a reference table for weakly informative priors for intercepts, coefficients, and scale parameters based on recommendations from experts like Andrew Gelman.

Is this skill suitable for count data?

Absolutely. It includes both Poisson regression for standard counts and Negative Binomial regression for scenarios where the variance exceeds the mean (overdispersion).

Bayesian Regression Models

Name: Bayesian Regression Models
Author: choxos

bychoxos

Data Science & ML

Implements diverse Bayesian regression techniques using Stan and JAGS for advanced statistical modeling and uncertainty quantification.

About

This skill equips Claude with specialized patterns and best practices for developing Bayesian regression models. It covers a wide range of data types, providing implementation code for linear, logistic, Poisson, and Negative Binomial regression. It also includes advanced techniques like robust regression using Student-t distributions to handle outliers and QR decomposition to optimize performance when dealing with correlated predictors. By providing curated prior recommendations and posterior predictive check patterns, this skill ensures that your probabilistic models are statistically sound and computationally efficient.

Key Features

Negative Binomial modeling for overdispersed count data
Robust regression patterns for handling dataset outliers
0 GitHub stars
Weakly informative prior recommendations based on industry standards
Stan and JAGS implementations for multiple regression types
QR decomposition templates for improved sampling efficiency

Use Cases

Analyzing count data in epidemiology or marketing where overdispersion is present
Optimizing predictive performance in datasets with highly correlated variables
Developing scientific research models that require rigorous uncertainty quantification

About

Key Features

Negative Binomial modeling for overdispersed count data
Robust regression patterns for handling dataset outliers
0 GitHub stars
Weakly informative prior recommendations based on industry standards
Stan and JAGS implementations for multiple regression types
QR decomposition templates for improved sampling efficiency

Use Cases

Analyzing count data in epidemiology or marketing where overdispersion is present
Optimizing predictive performance in datasets with highly correlated variables
Developing scientific research models that require rigorous uncertainty quantification