Weighting involves assigning weights to different cases in a data set when performing analysis. There are two common applications of weighting and multiple exotic applications/types.
The two common applications of weighting
Volumetric analysis
Volumetric weighting is where each case in the data is assigned a value indicating, in some sense, its worth, and then this weight is used to perform weighted calculations (see Example of a Weighted Calculation).
For example:
 When calculating market share, sales of products can be volumeweighted by their price.
 When calculating market share from a survey, data can be volumeweighted based on how much people consumed.
Sampling weights
Sampling weights have the goal of correcting for biases in how a sample has been collected. For example, if a population contains 49% men, but a sample for a survey contains only 26% men, the results can be weighted to correct for this underrepresentation.
A simple example of the creation and application of sampling weights is in Simple Worked Example of Creating and Applying a Sampling Weight. For more detail, see How to Weight a Survey.
Weighting a sample to deal with biases in data collection is also known as:

 Probability weighting
 Weighting survey data
 Sample balancing
 Calibration (which is technically a specific algorithm)
 Postsurvey adjustment
 Raking (which is technically a specific algorithm)
 Poststratification
 Nonresponse weighting
Exotic applications/types of weights
Although typically weighting refers to one of the above two types of weights, there are applications of the basic idea:
Comments
0 comments
Please sign in to leave a comment.