Modelling Neuronal Activity with Jittered Generalised Linear Models
MetadataShow full item record
By measuring electrophysiological data from a rat's brain we are able to study the relations between cells, and thereby study the brain itself. In this thesis we present different statistical modelling techniques for treating neuroscience data, detecting tuning of cells, detecting connectivity between cells and thereby investigating the flow of communication in the brain. We identify head direction and spatially tuning of cells with different neuroscience modelling techniques. Most importantly we develop the new method, jittered generalised linear models (JGLM). JGLM combines the best parts of the jittered cross-correlation method and the generalised linear model into one framework, and utilises permutation test on the jittered likelihood-values to test for connectivity between cells. JGLM is useful in detecting connectivity between cells and studying information in network of neurons. Additionally we develop the tool of basis-tuning-curve, which we use for classifying the type of connection between neurons. We have discovered that interval jittering is a good jittering procedure, while basic jittering is the wrong jittering choice, for jittering both in the JGLM and the JCC framework for analysing neuronal activity. The data analysed from the Oliva-16 data set contains electrophysiological data from cells recorded from movement related brain areas, as well as movement data for the rat. In the end we study the neuroscience findings from investigating the data set with JGLM. Furthermore we discuss the potential of JGLM and key elements in future research on JGLM, where ground-truth-testing of JGLM is essential. Our results shows that JGLM is well suited for analysing neuronal activities.