Analysing limited electrophysiological data by artificial sample generation

LUCAS MOLAS; CAMILO JUAN MININNI; SILVANO B. ZANUTTO; SERGIO E. LEW

Huerta Grande. Cordoba

Congreso; XXIX CONGRESO ANUAL DE LA SOCIEDAD ARGENTINA DE INVESTIGACION EN NEUROCIENCIAS; 2014

SOCIEDAD ARGENTINA DE INVESTIGACION EN NEUROCIENCIAS

P108.-Analysing limited electrophysiological data by artificial sample generation Lucas Molas1°, Camilo Juan Mininni2°, Silvano B. Zanutto1°, Sergio E. Lew1° 1° Instituto de Ingeniería Biomédica (IIBM-UBA), Facultad de Ingeniería, Universidad de Buenos Aires, Ciudad de Buenos Aires, Argentina; 2° Instituto de Biología y Medicina Experimental (IByME - CONICET), Laboratorio de Biología del Comportamiento, Ciudad de Buenos Aires, Argentina camilojm15@gmail.com During electrophysiological studies of animal behavior, subjects are required to achieve a certain (high) performance, in order to consider that they have learned a task. Given the finite number of trials an animal executes during a session before it gets satiated or exhausted, incorrect trials are a minority and statistically hard to analyze. When the purpose of a research is to understand the neural mechanisms that lead to an incorrect response, robust procedures to extract information from limited trials are required. In this work we propose a new method for the generation of samples which retain the statistics exhibited by real extracellular multielectrode recordings. The method compensates the bias produced by a reduced number of samples, and is tested for several statistical measures.