An Investigation into Adult Human Height Distributions Using Kernel Density Estimation

This study investigates how average adult human height distributions of various regions around the world have changed over time using a non-parametric approach. Performance kernel density estimators (KDEs) were compared between mixtures Gaussian created different means, variances and mixing weights. The performance was evaluated for these existing bandwidth selection methods, with kernels sample sizes it revealed distinct multi modes Sheather & Jones method performed better in general among considered. results this also that practical than can be achieved relatively smaller samples from gaussian through modified plug-in bandwidth. By applying findings simulation analysis on data related to heights cohorts, interesting observations made.