MAKING IT USEABLE    Weighting       Data Weighting Introduction

Weighting is the procedure to correct the distributions in the sample data to approximate those of the population from which it is drawn. This is partly a matter of expansion and partly a matter of correction or adjustment for both non response and non coverage. It serves the purpose of providing data that look like the population rather than like the sample.

Standard methods to develop weights rely on data from another source that show the distribution for the whole population. Weights were constructed for NPTS using estimates from the 1995 Current Population Survey of the U.S. Bureau of the Census.

IT IS ESSENTIAL TO USE WEIGHTED DATA FOR ALL ANALYSES because the unweighted data are not representative of anything other than the sample of households that responded.

Different weights are provided for different levels of analysis:

• Usable household level, which is based on the distribution of completed households in the sample and all households in the nation, and which takes account of differences between the two.
  
  
• Person level, which adjusts for person nonresponse within usable households by race, age, and sex.  These weights are used for all person-level data, and form the basis of the travel-day and the travel period weight.
  
  
• Travel-day level, which is based on those persons providing valid data for the travel day.
  
  
• Travel-period level, which is based on those persons who provided data on long-distance travel in the travel period.

The public use dataset includes the add-on samples.

The weights are adjusted so as to balance the add-on samples against the rest of the nation. Thus, low weights are given to some observations from the add-on samples, and higher weights to others, so that characteristics of the population, such as population size, gender, race, and geographic dispersion, remain representative of the entire nation, rather than looking like the add-on sample states and urban areas.

---