Enhanced Visual Statistical Inference: Comparative Evaluation with Linear Model Testing

This paper extends the framework of visual statistical inference, focusing on the lineup protocol as an alternative to conventional hypothesis testing. Through a series of human-subject experiments using simulated data, the study evaluates the lineup protocol’s effectiveness compared to traditional statistical tests in the context of linear models. Results show that the lineup protocol performs comparably to conventional tests and excels in situations where assumptions for traditional tests are violated, such as contaminated data scenarios. Notably, the lineup approach shows higher power for detecting effects when the effect size is large. The study also explores individual differences in visual inference ability, uncovering a subset of “super-visual” individuals who consistently outperform standard statistical methods. These findings suggest that visual inference, particularly the lineup protocol, offers a viable, and in some cases superior, alternative to traditional statistical testing, especially in exploratory data analysis contexts where conventional tests may be unavailable or unreliable.