Weighting adjusts your report data so it matches the ideal proportions of various groups in the general population or in a specific set of participants.
Use weighting to correct sampling or response rate issues.
Response rate correction example
You invite an equal number of men and women to complete your survey, and you find that more women than men completed the survey. You use weighting to correct the gender imbalance and see what responses would look like if the groups were balanced equally.
Weighting is most effective when your actual survey data proportions are close to your ideal proportions. In cases where the actual and ideal (or target) proportions are quite different, weighting is less advisable and your results may be less reliable. In certain cases, if the numbers are too divergent, it may not even be possible to calculate weights.
About Alida's weighting solution
The application uses random iterative method (RIM) weighting, otherwise known as raking, iterative proportional fitting, or iterative proportional weighting.
- Single Choice questions
- Single Choice Grid questions (rows)
- Single Choice profile variables
- Net Promoter Score℠1
The weighting algorithm does not perform any imputation on missing results, as this may distort data. We recommend filtering out participants with missing data before you begin.
Similarly, the weighting algorithm does not rescale weights in cases where your weight scheme may distort the influence of certain groups. We recommend adjusting your weights to reflect the demographics of the general population. Alternately, you can merge or unmerge population groups before you apply your weight scheme so you don't have small groups of outliers whose responses are given too much weight.
Weighting is applied when the weight scheme is first created, and refreshed when new responses are collected and the report data changes.
- You can apply a weight scheme to merged answers. If you merge answers after you apply the weight scheme, an error message will appear.
- To reduce the wait time loading weight variables, add weight variables with their original weight first, and then adjust the weighting.
Useful terms and definitions
The following terms and definitions appear in the documentation and application.
| Term | Definition |
|---|---|
| Weight scheme | The combination of weight variables, variable order, and targets that you specify for use in weighting. |
| Target percentage | The ideal percentage of a group within the whole, usually based on the actual percentage found in the general population or in a specific set of participants. |
| Sample percentage | The actual percentage of responses found in your data. |
| Weight variable |
A question or profile variable used in a weight scheme. |
| Answer category |
An answer to a question, or a profile variable value, used in a weight scheme. |
| Term | Definition |
|---|---|
| Random iterative method (RIM) weighting |
Also known as raking, iterative proportional fitting, or iterative proportional weighting. A method of weighting data using multiple independent variables. The RIM weighting algorithm attempts to calculate and balance target values for all variables in a weight scheme to achieve convergence (that is, to make the data match the targets set for all variables). |
| Iteration | One execution of the weighting algorithm across a set of data. The algorithm attempts to calculate and balance the target values for all variables in a weight scheme to achieve convergence. Weight schemes with more than one variable may require multiple iterations to achieve convergence. |
| Convergence | The successful calculation of a multi-variable weight scheme. Convergence indicates that the weight values for all the variables have been calculated and balanced, and that the weighted data matches the targets defined for all variables as closely as possible. |
| Weight factor (also known as weight value or weight) | The numeric weight value applied to each participant. |