Detection of dependence patterns with delay
Résumé
The Unitary Events (UE) method is a popular and efficient method used this last decade to detect dependence patterns of joint spike activity among simultaneously recorded neurons. The first introduced method is based on binned coincidence count (Grün, 1996) and can be applied on two or more simultaneously recorded neurons. This counting method is known to be subject to loss in synchrony detection (Grün et al., 1999). This defect has been corrected by the multiple shift coincidence count (Grün et al., 1999) for discrete time recordings of two simultaneously recorded neurons. A transposition to the continuous framework has recently been proposed in (Muiño and Borgelt, 2014). The statistical study in the continuous framework have been investigated in (Tuleau-Malot et al., 2014) for two neurons. The extension of this study to more than two neurons has not been performed until the present work. The point processes framework allows computations leading to a Gaussian approximation of the count for Poissonian spike trains. Since unknown parameters are involved in the approximation, a plug-in step is needed (where unknown parameters are replaced by estimated ones) and leads to a modification of the limit distribution. Finally we propose a multiple test procedure via a Benjamini and Hochberg approach (Benjamini and Hochberg, 1995). We compare our new method and the UE method proposed in (Grün et al., 2002) over various simulations including changes in the underlying model. Furthermore our method is applied on real data.
Origine : Fichiers produits par l'(les) auteur(s)