Creating Multiple Weights for Different Purposes

Sometimes multiple weights will be created for different purposes. For example:

• A weight may be created for the total sample including a boosted sub-sample and another which excludes the boosted sub-sample.
• Some analyses may have a different weight because they are known to be more biased by aspects of the data collection process.
• A volumetric weight may be created to consider differences between people in the amount they consume.
• One weight may be created for monthly analyses and another for quarterly analyses.
• Where analysis is conducted using rolling samples, a separate weight may be created for each rolling period (e.g., one weight for May to July, and another for June to August, etc.).

The last two of these should be avoided wherever possible. The reduction of bias achieved by such techniques will typically be negligible, while the increased difficulty of analysis and the resulting errors and time wasted will typically be non-negligible (e.g., there’s no rigorous way of comparing waves of a tracker that contain overlapping samples with different weights).