On Kernel-Based Mode Estimation Using Different Stratified Sampling Designs

In the literature, the properties and the application of mode estimation is considered under simple random sampling and ranked set sampling (RSS). We investigate some of the asymptotic properties of kernel density-based mode estimation using stratified simple random sampling (SSRS) and stratified ranked set sampling designs (SRSS). We demonstrate that kernel density-based mode estimation using SRSS and SSRS is consistent, asymptotically normally distributed and using SRSS has smaller variance than that under SSRS. Improved performance of the mode estimation using SRSS compared to SSRS is supported through a simulation study. We will illustrate the method by using biomarker data collected in China Health and Nutrition Survey data.