Raking (how to create a weight variable)
Raking (how to create a weight variable)
To create a weight variable for survey data, you will need three things:
1. A "raking model" to determine which variables will be used to construct the weight. Examples of these types of variables are county, race, sex, education level.
2. A dataset that represents the actual population of the region represented by your survey; for example, you can use census data, PUMS, or ACS.
3. A raking program, one can be found from the SUGI 25 proceedings paper "A SAS Macro for Balancing a Weighted Sample": http://www2.sas.com/proceedings/sugi25/25/st/25p258.pdf
You will need to generate frequencies of each variable in the raking model, using the census data. You will use these frequencies, along with your survey data itself, as input to the raking program. The raking program will add a weight variable to your dataset based on these variables.
Any observation with one or more missing values among the raking model variables will be assigned a weight of missing.
