Detecting when timeseries differ: Using the Bootstrapped Differences of Timeseries (BDOTS) to analyze Visual World...

Area of Science:

• Cognitive Science
• Psycholinguistics
• Computational Linguistics

Background:

• Advances in language sciences rely on real-time processing measures like mouse-tracking, event-related potentials, and eye-tracking.
• These methods generate dense timeseries data crucial for understanding cognitive dynamics.
• Current statistical methods for timeseries analysis have not kept pace with these data generation advancements.

Purpose of the Study:

• To introduce a new statistical approach, Bootstrapped Differences of Timeseries (BDOTS), for analyzing cognitive processing timeseries.
• To provide a method for precisely estimating the time window at which two timeseries diverge.
• To offer a flexible and robust statistical tool for psycholinguistic and cognitive science research.

Main Methods:

• The study presents the theoretical foundation of the Bootstrapped Differences of Timeseries (BDOTS) method.
• BDOTS makes minimal assumptions regarding error distributions.
• A custom family-wise error correction is incorporated, enhancing statistical rigor.

Main Results:

• The BDOTS method accurately estimates precise time windows of difference between timeseries.
• The approach is demonstrated through an analysis of an existing dataset, validating its practical application.
• The associated R package facilitates implementation and use.

Conclusions:

• BDOTS offers a significant advancement in statistical analysis for real-time cognitive and language processing data.
• The method's flexibility allows adaptation to diverse research applications.
• Recommendations for reporting BDOTS analyses are provided to ensure clarity and reproducibility.