A Method for Stationarity-Segmentation of Spike

1 Train Data with Application to the Pearson Cross2 Correlation 3 4 Claudio S. Quiroga-Lombard, Joachim Hass, Daniel Durstewitz 5 6 Bernstein Center for Computational Neuroscience, Psychiatry, Central Institute of Mental 7 Health, Medical Faculty Mannheim/ Heidelberg University, Mannheim, 68159, Germany 8 9 Address for reprint requests and other correspondence: Daniel Durstewitz, Central Institute of 10 Mental Health, Heidelberg University, E-mail: [email protected]. 11 12 Running Head: Stationarity-Segmentation of Spike Train Data 13 14 Abstract 15 Correlations among neurons are supposed to play an important role in neural 16 computation and information coding in the nervous system. Empirically, 17 functional interactions between neurons are most commonly assessed by 18 cross-correlation functions. Recent studies have suggested that pairwise 19 correlations may indeed be sufficient to capture most of the information present 20 in neural interactions. Many applications of correlation functions, however, 21 implicitly tend to assume that the underlying processes are stationary. This 22 assumption will usually fail for real neurons recorded in vivo since their activity 23 during behavioral tasks is heavily influenced by stimulus-, movement-, or 24 cognition-related processes, as well as by more general processes like slow 25 oscillations or changes in states of alertness. To address the problem of non26 stationarity, we introduce a method for assessing stationarity empirically and 27 then “slicing” spike trains into stationary segments according to the statistical 28 definition of weak-sense stationarity. We examine pairwise Pearson cross29 Articles in PresS. J Neurophysiol (May 1, 2013). doi:10.1152/jn.00186.2013