Comparing weighted and non-weighted anthropometric and categorical data I wish to do a comparative analysis between a primary, local data set and the National Health and Nutrition Examination Survey (NHANES).  Specifically, I'd like the comparison to be limited to 11 year olds (my survey was only sixth graders, hence 11 year olds) and the intent here is to compare means for four anthropmetric variables and more importantly to compare obesity prevalence based on category frequencies between the two data sets (normal, overweight, obese). 
However, my dataset is unweighted (we measured at least 80% of the 6th grade population in the area), while the NHANES is weighted (its sampling design involves oversampling under-represented populations).  I have been attempting to conduct my analysis in SAS, as it is what is most recommended by the NCHS (National Center for Health Statistics) for working with the NHANES.  
Any suggestions on how I should proceed from here? 
Here is a link to the Continuous NHANES tutorial in case more information on the sampling design is needed: http://www.cdc.gov/nchs/tutorials/Nhanes/SurveyDesign/SampleDesign/intro.htm
 A: The NHANES website is quite thorough. Have you adequately reviewed their examples and data descriptions? Here's an example for how they apply their survey weights:
http://www.cdc.gov/nchs/tutorials/NHANES/NHANESAnalyses/DescriptiveStatistics/Task3b.htm
In their breakdown of the PROC SURVEYMEANS, they show this line of code:
weight wtmec4yr;
with the following description

Use the weight statement to account for the unequal probability of sampling and non-response. In this example, the MEC weight for four years of data (wtmec4yr) is used.

As you may be aware, survey weights are analytic weights intended to account for non-response and sampling strategies to reliably ascertain trends in rare populations. 
A: a few notes about what you've described  :)
1) AdamO is correct that once you calculate the statistic of interest from the weighted NHANES data set, that's the survey's estimate of the population..and if you have a comparison statistic also at the population level, you can just say "yes" or "no" RE whether your MSA's results are inside the nhanes confidence interval.  that is: "the BMI of sixth graders in our MSA are/are not statistically significantly different from the average as computed from a nationwide sample."
here's some easy-to-use scripts to work with nhanes in the R language
2) if BMI is your most important item, doing what you've described is much better done with the "sample child" component of the national health interview survey than with nhanes.  the two-year nhanes file has about 200 records for 11 year olds.  using the four-year weights means you've got a weird sample population of 400 11 year olds across the survey years 2007-2010.  the 2012 nhis file is already available, and has 600+ respondents who are 11 years old.
here's the page with easy-to-use scripts to work with nhis ..and you specifically should modify this script to work with the samchild (instead of samadult) file on the 2012 data sets.
