Statistical analysis and modeling of neural data Lecture 7 Bijan Pesaran 26 Sept, 2007
Goals Recap – spectral analysis for point processes. Coherency/Partial coherency as linear measures of association. Time-rescaling theorem and model validation
Spectral analysis for point processes Regression for temporal sequences Naturally leads to measures of correlation Statistical properties of estimators well- behaved
Cross-spectral density
Spectral quantities
Coherence as linear association
Minimum value is: Where: Minimize wrt B(f): Substitute into loss:
Time lags in coherency
Partial coherence
Time-rescaling theorem Poisson with rate 1 History-dependent time rescaling
Kolmogorov-Smirnov Test Empirical distribution function Critical values of the Kolmogorov-Smirnov Distribution
Kolmogorov-Smirnov Test Test for deviation from uniform distribution Construct cumulative distribution function Rank vs (95% confidence)(99% confidence)