Variance and Distribution Models for Steering Tasks.

Area of Science:

• Human-Computer Interaction
• Cognitive Psychology
• Motor Control

Background:

• The steering law describes a linear relationship between movement time (MT) and index of difficulty (ID) in steering tasks.
• Existing models do not account for the variance or distribution of MT in relation to ID.

Purpose of the Study:

• To propose and evaluate models for predicting the variance and distribution of MT based on ID for steering tasks.
• To advance MT prediction from point estimates to variance and distribution estimates.

Main Methods:

• Proposed a quadratic variance model relating MT variance to ID.
• Empirically evaluated the quadratic variance model against linear and constant models using new and existing datasets.
• Assessed six distribution types (e.g., Gamma, Lognormal, Gaussian) for predicting MT distribution given ID.

Main Results:

• The quadratic variance model explained 78-97% of the observed MT variance, outperforming other models.
• Positively skewed distributions (Gamma, Lognormal, ExGaussian, Extreme Value) outperformed symmetric distributions (Gaussian, truncated Gaussian) for MT distribution prediction.
• Gamma distribution showed slightly superior performance among the skewed distributions.

Conclusions:

• The quadratic variance model accurately predicts MT variance in steering tasks.
• Incorporating variance and distribution predictions offers a more comprehensive understanding of steering behavior and quantifies MT prediction uncertainty.