Power law versus exponential state transition dynamics: application to sleep-wake architecture

Area of Science:

• Neuroscience
• Computational Biology
• Sleep Science

Background:

• Interrupted sleep negatively impacts waking function, yet sleep architecture features distinguishing continuity from fragmentation remain unclear.
• Characterizing sleep architecture via temporal dynamics of sleep-wake stage transitions is an area of growing interest.
• Sleep and wake bout durations are modeled by exponential and power law distributions, respectively, but distinguishing these can be complex.

Purpose of the Study:

• To investigate the parameters that cause multi-exponential and power law distributions to mimic each other.
• To understand the implications of model mimicry on sleep architecture analysis.

Main Methods:

• Systematically fitted multi-exponential distributions with a power law model and vice versa.
• Employed the Kolmogorov-Smirnov method to assess goodness of fit for incorrect models across various parameters.
• Identified a "zone of mimicry" where parameter choices increase the risk of misclassifying distributions.

Main Results:

• Multi-exponential distributions can resemble power law distributions, appearing linear on log-log plots.
• The identified "zone of mimicry" parameters align with empirical time constants from human sleep and wake bout distributions.
• This mimicry highlights the difficulty in definitively distinguishing between exponential and power law models for sleep-wake dynamics.

Conclusions:

• Uncertainty in distinguishing between exponential and power law models impacts the interpretation of sleep-wake transition dynamics (e.g., self-organizing vs. probabilistic).
• This ambiguity affects the development of predictive models for classifying normal and pathological sleep architecture.
• Accurate model distinction is crucial for understanding fundamental sleep processes and for clinical applications.