Regression Confessions: Stories of Statistical MisbehaviorJanuary 21, 2026 •4:00 PM UTC

Regression is a powerful tool for uncovering relationships in your data. But with great power comes… well, plenty of opportunities to misuse it. Many of the most common regression analysis mistakes happen not out of carelessness, but because the results look convincing and we don't dig deep enough to answer the question, “Can I really trust this?” As George Box famously warned us, “All models are wrong, but some are useful.” The challenge is knowing when your model is genuinely useful… and when it’s quietly leading you astray.

In this webinar, we’ll break down real examples of regression analysis gone wrong: overfit models, misleading p-values, suspiciously perfect R² values, spurious correlations, and other statistical shenanigans that would make George Box shake his head. You’ll learn what these mistakes look like, how to avoid them, and how to use regression analysis as a force for good in your work rather than a source of accidental mischief.

This session is designed for continuous improvement practitioners, analysts, and leaders who want to sharpen their instincts, strengthen their analysis, and confidently separate meaningful insights from statistical misbehavior.

What You’ll Learn

By the end of this session, participants will be able to:

• Identify common forms of regression misuse, including overfitting, p-hacking, ignoring model assumptions, and treating artifacts as insights.
• Recognize misleading interpretations such as misused p-values, inflated R² values, and spurious correlations that can derail decisions.
• Explain statistical vs. practical significance and why both matter when recommending action.
• Assess data quality and variable selection choices that can bias or distort results.
• Interpret real-world examples of regression “misbehavior”