The Incidental Parameter Problem in Network Analysis for Neural Spiking Data Nadkarni, Dahlia creator 2015 xii, 137 p. born digital Thesis (Ph.D. -- Brown University (2015) Harrison, Matthew Director Geman, Stuart Reader Amarasingham, Asohan Reader Brown University. Applied Mathematics sponsor theses The spiking electrical activity of simultaneously recorded neurons is often modeled as a multivariate binary time series. Inference about the fast temporal correlation structure of this multivariate time series, such as zero-lag synchrony and precise lag-lead relationships, is complicated by the time varying response of the neurons to their many unobserved and correlated inputs. Classical approaches to this problem suffer from model misspecification or incidental parameter problems in which the number of nuisance parameters grows with the size of the data. We develop a conditional inference approach that works well for inferring the synchrony parameters in log-linear models (maximum entropy models) with non-stationary background firing rates. We also explore a composite likelihood (or the pseudo-likelihood) approach that scales better to larger numbers of neurons and more complex models, and an alternate Bayesian non-parametric approach using Dirichlet process mixture (DPM) priors for the nuisance parameters. Incidental parameter problems Conditional inference RPB 20150601 engEnglish10.7301/Z04F1P47In CopyrightCollection is open for research.dissertations