"Density Estimation Through Bayesian Regression Splines"

Abstract: Consider this situation: you have been given a set of data x=x1, x2,…,xn to analyze, but you only have very limited information about this data. All you know about this data is that it comes from some probability density. With only this information, where would you begin? You might be able to make a conjecture of the identity of the true density function from which the data came, by plotting the data and comparing its shape to that of a known parametric density. However, what if your data did not come from any such function? Then how would you begin to analyze it? You could use a smoothing technique to approximate the true density from which the data came. In this talk, I will introduce a method for estimating this density using regression splines in a Bayesian statistical framework and demonstrate how this method outperforms many of the common curve-fitting techniques employed today.