MAKING IT USEABLE    Weighting       Weights Related to Sample Strata

Weighted data will not necessarily produce correct results for all subsets of the dataset:

• When subsets of the data are used, population and other figures will not reflect accurately those of the sub-areas contained. For example:
  
  
  • The correct population of Louisiana is 4,350,579; while abstracting the Louisiana observations and applying household weights gives a population for the state of 3,596,583.
    
    
  • In the same analysis, the population of New Orleans appears to be 1,040,550 compared to the true value of 1,315,294; and for Baton Rouge, the figure from the subset of the NPTS is 498,138 where the correct figure is 563,994.

• The weights are designed to replicate the population of the country and certain characteristics for the entire country. Any time that a subset of data is analyzed, the weights will not replicate the exact distributions and totals of that subset, unless the subset coincides exactly with the strata of the survey design as discussed in Module 1.

---