Weighting data (respondent weighting)
Respondent weights easily added to data
Rim or matrix, using single and multi-coded entries
Targets as values and/or percentages
Constraints on weight values
Weighting report automatically generated
Respondent weighting is the process of allocating a weight to each individual record (questionnaire) to rebalance analysis including:
- Matching known population sizes
- Removing the effects of deliberate over/under sampling of a section of the respondents
- To compensate for failure to achieve exact quotas
- To equalize demographics for repeated/continuous surveys
A weight is stored in a float (floating point) entry in every record and for analysis this value is added instead of 1.0 whenever this record enters a table or report.
The desired weighted total value for each subset of the data is called a “target” which can be a figure or a percentage, hence the term “target weighting”.
A set responses to an entry with targets is called a “rim”, hence the term “rim weighting”.
The sum of the targets for a rim will usually be 100%.
QPSMR Companion has a weight procedure which you can use to allocate weights to the individual data records.
There is no limit on the number of rims and multi-coded entries you can use. The outline procedure is usually:
- Use [Data] [Weight data] to open the Weight data window
- Choose the data file to weight
- Choose or create the float entry to store the weights
- Set and save the targets
- Prepare the data and calculate the weights
- Check the summary report, especially the ESS
- Save the weighting summary report
- Then save the data with the weights
- You should then save the project (QDF) if you create a new weight entry
Important: You must inform whoever devised the weighting scheme of the overall effective sample size (ESS) which will be the base sizes for statistics. This is often shown as a “weighting efficiency” so that they can decide if the loss of data due to weighting effects is reasonable, or whether they need to simplify the targets to keep more of the data. The overall “weighting efficiency” is the ESS expressed as a percentage of the unweighted number of respondents.
It is not unknown for clients to devise comprehensive targets that “throw away” half the data collected through weighting effects. Where practical, you should save the weighting summary report with the project and sent with any weighted analysis.
Problems with weighting
Sometimes there will be a problem with the targets or the data which can result in:
- The program failing to achieve the requested targets
- The calculations take a very long time
Using with other software
You can use the weighting procedure to weight data which is processed by other software, for example: