Using stratification to estimate multimodal density functions with applications to regression

Abstract

Large populations, particularly large human populations, are usually composed of several different subpopulations. We want to utilize this fact to provide a means of improving on the choice of the bandwidth of the kernel density by combining the separate density estimates of the subpopulations. This will provide an improvement on the overall density estimate for the entire population.

Keywords:

• Stratified Sampling
• Kernel Density Estimator