hw_H: KDE bandwidth for residual variability in height/weight
In httk: High-Throughput Toxicokinetics

hw_H	R Documentation

KDE bandwidth for residual variability in height/weight

Description

Bandwidths used for a two-dimensional kernel density estimation of the joint distribution of residual errors around smoothing spline fits of height vs. age and weight vs. age for NHANES respondents in each of ten combinations of sex and race/ethnicity categories.

Usage

hw_H

Format

A named list with 10 elements, each a matrix with 2 rows and 2 columns. Each list element corresponds to, and is named for, one combination of NHANES sex categories (Male and Female) and NHANES race/ethnicity categories (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, and Other).

Details

Each matrix is a variance-covariance matrix for a two-dimensional normal distribution: this is the bandwidth to be used for a two-dimensional kernel density estimation (KDE) (using a two-dimensional normal kernel) of the joint distribution of residual errors around smoothing spline fits of height vs. age and weight vs. age for NHANES respondents in the specified sex and race/ethnicity category. Optimal bandwidths were pre-calculated by doing the smoothing spline fits, getting the residuals, then calling kde on the residuals (which calls Hpi to compute the plug-in bandwidth).

Used by HTTK-Pop only in "virtual individuals" mode (i.e. httkpop_generate with method = "v"), in gen_height_weight.