Statistical Analysis Using Large Language Models: Guidelines and Best Practice

Large Language Models (LLMs), such as ChatGPT, Claude, and Gemini, are increasingly used to support statistical analysis. They can help researchers understand statistical methods, generate R code, interpret results, and improve scientific reporting. However, incorrect statistical recommendations, inappropriate model selection, and misleading interpretations can seriously affect research quality.

This two-day interactive workshop introduces practical guidelines for using LLMs safely, effectively, and responsibly in statistical analysis.

Participants will learn how to communicate their research questions and study designs to LLMs, evaluate suggested statistical methods, generate and verify statistical code, assess model assumptions, and interpret results appropriately.

The workshop combines short presentations, real research examples, and hands-on exercises, following the SAFE-STAT framework developed in the Statistics@SLU guidelines.

Day 1 (Nov 11; 13:15-16:00):

Introduction to LLMs in statistics; common mistakes; effective prompting; understanding study design and variables; choosing appropriate statistical methods; practical examples

Day 2 (Nov 12; 13:15-16:00)

Generating and debugging R code; checking assumptions and diagnostics; interpreting statistical results; detecting LLM errors; verification, reproducibility, and data privacy; hands-on exercises
Questions, Answers and Discussions. 

Leader: Reza Belaghi,  Department of Energy and Technology, Unit of Applied Statistics and Mathematics, SLU