Ming-Yen Cheng

Estimating bivariate density and regression surfaces

We consider nonparametric estimation of
bivariate surfaces of density and regression functions. The original
data are first properly smoothed or processed, in a way that the
results retain useful information, and then applied to simple conventional
nonparametric methods. The first step generates pseudo data points
larger in amount than the original data, and has the potential to
reduce difficulty caused by data sparseness problems. Theoretical and
numerical properties of the techniques are examined.