Online Program

Saturday, February 21
CS23 Population Modeling Sat, Feb 21, 11:00 AM - 12:30 PM
Napoleon C

Determining Different Population Distributions Using NHANES BMI Data (302921)

Robbie A. Beyl, Pennington Biomedical Research Center 
Jeff Burton, Pennington Biomedical Research Center 
*William Johnson, Pennington Biomedical Research Center 

Keywords: NHANES, percentiles, KS test, non-parametric

A common question when dealing with NHANES (National Health and Nutrition Examination Survey) data is, “Do different populations (e.g., race and sex) have the same probability distribution?” Nonparametric tests, such as the Wilcoxon rank sum test, are based on the location parameter. Thus, these tests are less powerful when comparing samples from distributions with similar location parameters but vastly different shape parameters. We propose an alternative test, based on the empirical distribution functions, that uses percentiles to characterize the distributions. The percentiles are used to categorize the data, and a chi-squared test statistic is constructed. Using simulated data, empirical properties of the test (size and power) are shown under various settings. These properties are compared to other similar empirical-based tests such as Kolmogorov-Smirnov and Kuiper’s, as well as the Wilcoxon test. The NHANES data on BMI is used to showcase how our test is able to detect differences in certain sub-populations.