The time-rescaling theorem and its application to neural spike train data analysis

Area of Science:

• Computational Neuroscience
• Probability Theory
• Statistical Modeling

Background:

• Evaluating the goodness-of-fit for neural spike train models is essential for validating statistical models before inferring neural system behavior.
• Assessing goodness-of-fit is particularly challenging for histogram-based models like per-stimulus time histograms (PSTH) and smoothed rate functions.
• The time-rescaling theorem provides a theoretical basis for transforming point processes into unit-rate Poisson processes.

Purpose of the Study:

• To adapt the time-rescaling theorem for developing goodness-of-fit tests for neural spike train models.
• To create a unified paradigm for comparing parametric and histogram-based point process models.
• To enhance the accessibility of the time-rescaling theorem for neuroscience researchers.

Main Methods:

• Applied the time-rescaling theorem to develop goodness-of-fit tests for parametric and histogram-based point process models.
• Utilized elementary probability theory to present a proof of the time-rescaling theorem.
• Demonstrated the theorem's utility in simulating spike train point process models.
• Compared PSTH, inhomogeneous Poisson, and inhomogeneous Markov interval models for macaque monkey data.
• Compared temporal and spatial smoothers, inhomogeneous Poisson, gamma, and inverse Gaussian models for rat hippocampal data.

Main Results:

• The time-rescaling theorem provides a viable method for assessing goodness-of-fit in neural spike train models.
• The developed tests enable direct comparison of parametric and histogram-based models.
• The theorem facilitates the simulation of general point process models for spike trains.
• The study successfully applied these tests to real neural data from monkeys and rats.

Conclusions:

• The time-rescaling theorem is a valuable and versatile tool for analyzing neural spike train data.
• This approach simplifies the evaluation and comparison of diverse neural spike train models.
• The findings suggest broader applicability of the time-rescaling theorem in computational neuroscience.