|
Abstract:
|
Beginning with the 2011-2012 cycle, NHANES drew a special one-third subsample of adults comprising all current tobacco smokers and a random selection of non-smokers from Subsample A. This Smoking Subsample was intended to augment biomonitoring of smokers by essentially oversampling this group, but many of the same biomarkers were measured in the Smoking Subsample as in Subsample A. Combining these subsamples may enhance overall statistical power, and application of full-sample weights would enable sample-weighted analysis. Using NHANES 2011-2014 urine concentration data for thiocyanate (N=8096), a tobacco smoke biomarker, this study assessed the performance of using full-sample weights for analysis of combined subsamples nested within the full sample. Sample-weighted variances estimated with the bootstrap (Rust & Rao 1996; 1000 replicates) for the sum of urinary thiocyanate found confidence limits of the combined subsamples improved between 28-36 percent over the subsamples individually. Commensurate improvements were observed for the sample-weighted mean and median, as well as the univariate regression slope for thiocyanate's association with urine flow.
|
