Nonparametric density estimation for bounded data

This thesis proposes nonparametric methods for the estimation of univariate density in a unit interval, as well as bivariate copula density. A unified Bayesian framework is proposed for choosing the optimal kernel function and bandwidth. The results of the simulation studies establish the finite sample performances of proposed methods relative to existing counterparts. Relative to existing methods, the proposed methods are shown to adequately estimate the univariate probability density of recovery rates of financially distressed firms and bond issuers. Additionally, the proposed copula has superior tail risk forecasts of All Ordinaries index returns conditional on S&P 500 index returns.