For several NHANES files, data were collected or processed on a smaller sample (subsample) of individuals.
Each of these subsamples is a nationally representative sample.
For each subsample, the analytic sample size is smaller and appropriate sample weights have been calculated to reflect both the additional stage of sampling and the additional nonresponse (if any) for the subsample.
These weights differ from the full examination weight and the revised weights must be used for statistical estimation.
For each data file, the appropriate sample weights are included on the file along with the appropriate replicate weights needed for sampling error estimation.
A common processing error is to use either the wrong sampling weight or to use the wrong set of replicate weights.
Due to smaller sample sizes, the sampling errors for subsamples will be larger than for comparable data items on the full examination sample.
Also, the potential impact of outliers (large data values especially in conjunction with large sampling weights) should be addressed in any analysis of these data – especially when the prevalence is less than 10 percent.
Please note that when merging full sample data items to the subsample data files, the analytic sample is the subsample and the subsample weights must be used.
Users are strongly urged to read the data file documentation and to take care to ensure proper analysis and interpretation of the data.