Unbiased estimation of precise temporal correlations between spike trains

Area of Science:

• Systems Neuroscience
• Computational Neuroscience
• Neural Information Processing

Background:

• Precise temporal inter-neuronal interactions are crucial for brain information processing.
• Cross-correlation histograms (CCHs) are commonly used to study pair-wise neuronal interactions.
• Interpreting CCHs is challenging due to confounding factors beyond temporal correlations.

Purpose of the Study:

• To evaluate the validity of the jittering method for analyzing precise temporal interactions in neural spike trains.
• To develop and validate a novel, unbiased method for detecting temporal correlations between spike trains.

Main Methods:

• Analysis of the mathematical underpinnings of the jittering method, revealing its equivalence to convolution with a finite window and Poisson probability estimation.
• Development of a modified convolution method using a partially hollowed window.
• Validation of the new method using both artificial and real neural spike train data.

Main Results:

• The jittering method over-fits spike train data, resulting in biased statistical tests with low power.
• The modified convolution method is demonstrated to be unbiased and possess high statistical power.
• The modified convolution method is computationally efficient.

Conclusions:

• The jittering method and its interpretations should be approached with caution due to inherent biases.
• The modified convolution method provides a reliable and powerful tool for detecting precise temporal correlations in neural spike trains.
• This work offers improved analytical techniques for systems neuroscience research.