Using piecewise regression to identify biological phenomena in biotelemetry datasets.

Area of Science:

• Ecology
• Animal Behavior
• Data Science

Background:

• Technological advances in animal tracking provide rich biotelemetry data streams.
• Understanding animal responses to stimuli requires analyzing behavioral and physiological shifts over time.

Purpose of the Study:

• To demonstrate the utility of piecewise regression for detecting changes in animal biotelemetry data.
• To infer biological latent states by segmenting time-series data based on response variations.

Main Methods:

• Utilized piecewise regression to partition time-series data into segments with distinct model structures.
• Applied the mcp package in R for Bayesian analysis, model fitting, and assessment using information criteria and cross-validation.
• Demonstrated the approach with six diverse case studies across various species and data types.

Main Results:

• Successfully identified behavioral responses (e.g., flee and return), parturition, physiological stress responses, mortality events, migration patterns, and hatching events.
• Showcased the flexibility of piecewise regression in analyzing diverse biotelemetry data, including location, movement, heart rate, temperature, and acceleration.
• Highlighted the method's ability to infer distinct biological states and transitions.

Conclusions:

• Piecewise regression offers a robust and accessible method for analyzing biotelemetry data to detect biologically relevant changes in animal behavior and physiology.
• The approach facilitates the study of animal responses to environmental stimuli and internal states across various species and timescales.
• The mcp R package provides a user-friendly platform for implementing these advanced statistical analyses in ecological research.