|
|
 |
 |
 |
 |
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
INTRODUCTION: The 1999-2004 Dual Energy X-ray Absorptiometry (DXA) Multiple Imputation Data Files and Technical Documentation Typically, NHANES data sets with missing data/variables are released without statistical adjustment because the amount of missing data is small and the data items are usually missing at random. However, examination of missing items in the DXA data files indicated that the amount of missing DXA data was somewhat larger than for other data files and there seemed to be systematic, non-random patterns to the missing data. Use of only the measured variables could lead to biased results. To provide the user with a “complete” data file, missing and invalid DXA data from NHANES 1999-2004 were imputed using multiple-imputation methodology. Data files for 1999-2000, 2001-2002, and 2003-2004 can be downloaded via the Data links below. Each of the data files contains FIVE sets of measured and imputed values. Each set of measured and imputed values can be merged with other data from NHANES to create analytic datasets. NOTE: The multiple imputation procedure produced a small number of imputed values with extreme variability for some survey participants. Because of the extreme variability of these imputed values, the data for these participants have been placed in separate files labeled Supplemental Highly Variable Data. Analysts should be aware of the highly variable nature of these imputed DXA data when considering the use of these separate files. Multiple imputation is a technique that allows analysts to incorporate the extra variability due to imputation into their analyses. Imputed values should not be treated as measured variables without accounting for the extra variability introduced by the imputation process. The extra variability due to imputation CANNOT be incorporated by simply analyzing a SINGLE dataset as if the imputed values were true values. Moreover, analysts SHOULD NOT create a single dataset using the AVERAGE of the five sets of valid and imputed values. The preferred statistical approach is to analyze EACH OF THE FIVE datasets separately using methods and software that are appropriate for survey data and then combining the estimates and standard errors using the combining rules described in Section 4 of the document available via the Technical Documentation for Multiple Imputation link below. Section 4 of the Technical Documentation also contains examples for multiply imputed data analyses using SAS-callable SUDAAN. Answers to questions about the multiply imputed DXA data are available via the Frequently Asked Questions for the Multiple Imputation link. Information on the DXA examination, the data files (variable descriptions, file structure, control counts), and codebook are available via the Data File Documentation links below. Users interested in analyzing data for several years should note that the variables and file structure are identical for all three release cycles. ANALYSTS ARE STRONGLY ENCOURAGED TO READ BOTH THE TECHNICAL DOCUMENTATION AND THE DATA FILE DOCUMENTATION. Users are also encouraged to check the NHANES What’s New website for updates and to subscribe to the NHANES Listserv to receive notices of any corrections/updates.
2003-2004 2001-2002 1999-2000 Supplemental
This page last reviewed
December 22, 2008
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
