Topics in computational statistics with applications in finance



Journal Title

Journal ISSN

Volume Title



The dissertation comprises two parts, each commenced with an introduction to the principal ideas and methods therein: The first part concerns topics related to marginal and intractable likelihood estimation, focusing on the estimation of a density function at a particular point. In particular, we present a Monte Carlo estimator based on the Fourier integral theorem. The second part concerns Bayesian approaches to state-space models with applications in finance. First, we introduce a Bayesian vector autoregression to examine strategic asset allocation for long-run investors, given estimation risk and the choice of multiple risky assets. Then, we devise a variation on the Bayesian additive regression trees (BART) framework to incorporate time-dependent data, as well as stochastic volatility (SV), before applying this approach to the problem of predicting a firm’s stock return with observable firm characteristics. Joining the two parts is an interlude, which describes an approach to the particle filtering of hidden Markov models which reverses the standard sampling-resampling perspective and, along with several simulation studies, includes an example involving an SV model.



LCSH Subject Headings