The weight variables are created and maintained in the Weight screen located in the Reporting Objects Tab in the administration part.
When creating a Weight variable you can choose which method to use:
- Standard target weighting
- R.I.M weighing
The interface is divided into three sections:
- Select variables
- Define Target Distribution
The following settings are made in the Setting section:
Code and Name: each weight variable needs a code and name, the code must be unique.
Method: Select which weighting method to be used; Standard target weighting or R.I.M weighting
Weight within: Select what base to use for the weighting, the weights can be made within all data in the project, per imported data batch or within a certain interval like week, month, quarter or year.
Note: If, for example, Year is selected and new data is imported on a monthly level it means that the weight of the already imported respondents will be updated when data for a new month is imported as all respondents belonging to the same year is used as base in the weight calculations.
Target tolerance: The Target tolerance is used to validate so the sum of the defined targets is close to100%, in case of too big difference you will not be able to save the weight variable until you correct the targets (or update the Target tolerance), see an example below.
In this example the Target tolerance has been set to 0.5% which means that the sum of the targets must be between 99.5% to 100.5%, as the sum in the Age group variable is 101% the sum cell becomes red and you will not be able to save until the targets are corrected.
Rescale Target values during the weight calculation, if respondent count is zero: This option is used to handle edge cases where the respondent count is zero which could be the case in for example a tracker where the data is weighted within each week and the number of interviews are not big enough.
The defined Targets are as the following:
- Option A: 50%
- Option B: 30%
- Option C: 20%
If the actual respondent count of option B is zero during the weight calculation the Targets for the options A and C will be rescaled automatically during the weight calculation to sum up to 100% to not get a skew base size. The new target will be become:
- Option A: 71.43% (50/(0.5+0.2)
- Option B:
- Option C: 28.57% (20/(0.5+0.2)
Note: in the example above the new targets are shown with 2 decimals but 15 decimals are used in the actual calculations.
In the Select Variables section you select which variables to be used. Besides this you can have a filter variable applied in case of having the need of for example defining different targets for different countries.
The variable list contains all the single choice variables in the project, both the imported variables and computed variables. To select a variable hover over the desired variable and click “Add to weight table” or “Add as Filter”.
The image below shows how to add a variable to the weight table or as a filter.
Here we see an example where Age and Gender have been added to the weight table.
In case you added Country as a filter it would look like this:
To remove an already selected variable click the “Remove” option either in the variable list or in the weight table.
The image below shows how to remove a selected variable.
Define target distributions
In the Define Target distribution section the target distributions are defined in the weight table. The“sum” cell indicates if the sum of the targets is within the defined tolerance from 100% which it has to be to be able to save the weight variable.
The image shows the weight table, the “Current (%)” shows the actual unweighted distribution based onall the data in the project, the “Target (%)” column is used to enter the desired targets.
The option “Rescale Target values if the sum is not equal to 100%” can be used to rescale the targets automatically so these become 100%. If this is option is ticked you will be able to save the variable even if the sum is not within the defined tolerance from 100% as the new rescaled values will be used.
Here you see an example of when the “Rescale Target values if the sum is not equal to 100%” is ticked, the new rescaled targets are shown in a new column and these will be inserted as targets on save.
Note: if new answer categories are added during the project and the weight definition has not been adjusted these weight numbers will set as missing, counted as 1.
Note: Dapresy accepts respondents to fall in multiple categories, even though the question type is a single response. Because of this you can theoretically use these questions in the weight calculation, but this is not officially supported by Dapresy. (The weight for the respondents who fall in different categories is an average of the weight of the categories when you would do this with a Target weight).