Statistical assessment of time-varying dependency between two neurons

Area of Science:

• Neuroscience
• Computational Neuroscience
• Statistical Modeling

Background:

• The joint peristimulus time histogram (JPSTH) visualizes correlated neuronal activity but requires adjustments for time-varying firing rates.
• Existing methods for measuring time-varying correlated activity can be complex and may not efficiently handle firing rate modulations.

Purpose of the Study:

• To develop a statistical model for time-varying joint spiking activity to improve the efficiency of joint firing rate estimation.
• To introduce a robust statistical significance test for detecting correlated neuronal activity.

Main Methods:

• Applied an adaptive smoothing method to the ratio of joint firing probability to independence-predicted probability.
• Utilized a bootstrap procedure for a significance test, applicable to both Poisson and non-Poisson data.
• Conducted numerical simulations to evaluate the performance of the bootstrap-based significance test.

Main Results:

• The bootstrap-based significance test demonstrates high accuracy in rejection probability.
• This new method shows significantly better power in detecting departures from independence compared to testing contiguous bins in the JPSTH.
• The statistical model efficiently estimates time-varying joint firing rates.

Conclusions:

• The proposed statistical model and bootstrap significance test provide an efficient and powerful approach for analyzing time-varying correlated neuronal activity.
• This formulation is a valuable tool for understanding neural coding and network dynamics.
• The method is robust and adaptable for various neural data types.