Note: Data weighting is only available for Expert users
When to apply data weighting?
Data weighting is particularly useful in situations where certain groups are overrepresented or underrepresented in your initial sample composition. By applying weights post-data collection, you can correct for any biases that may have occurred during the data collection process. This allows you to work with representative data.
Data weighting is usually applied to demographic characteristics such as age and gender. These fields will be used in the example below to match the sample data to the Dutch population.
Representation demographic characteristics
| Survey Sample | Dutch Population | |
| Male | 70% | 47% |
| Female | 25% | 50% |
| Divers | 5% | 3% |
| Survey Sample | Dutch Population | |
| < 20 years old | 10% | 21% |
| 20 – 39 years old | 20% | 26% |
| 40 – 64 years old | 50% | 33% |
| 65 – 79 years old | 18% | 15% |
| > 80 years old | 2% | 5% |
In this example there is an overrepresentation of male and divers respondents and an underrepresentation of young (<39 years old) and elderly (>65 years old) respondents.
Download weighting data definition template in order to follow with the next steps on this manual
How to apply data weighting?
To make sure your collected data is a good representation of the Dutch population the desired weight is uploaded using an Excel file (example file is attached below). The file always contains three columns.
A. Preparing the Excel File
1.Variable
This column contains the variable names that you defined in your survey:
Note: In order to work with custom variables, “custom” prefix must be added in weighting template, for example, custom.age.
Adding member based variable is not directly possible, only if these are referred to with a custom variable.
2.VariableValue
This column contains the values of the variables in your survey:
3.Weight
This column contains the values from the age distribution column of the Dutch population.
Note: Only use the integer number excluding the percentage symbol (%):
In this example where two demographic characteristics (age & gender) are used this results in the following Excel format:
B. Uploading the Excel file
Navigate to the ‘Analyze’ tab of your survey and check the ‘Interviews weighting’ box to upload your Excel file:
Your Excel file will be validated based on the following:
- Weight can be specified with integer numbers or real numbers;
- The variable name must exist;
- If the variable is based on a question with a defined range of values (e.g. single response question) the defined value must be in the range;
- The weights for one variable must sum up to 100.
If the Excel file is validated you can determine whether you want to:
- Automatically calculate weights when the survey is closed. Here closing the survey triggers the weights to be calculated
- Manually decide when to calculate weights. Here you can manually trigger the calculation of weights, for instance on an active survey.
C. Analyzing the results
When the survey is closed you can analyze the results. The analysis of weighted interviews is only possible in Dashboard Reports.
In order to created tables with weighted data, go to Dashboard Report > Tables > Add content > Table Wizard, and select “interview weighting” setting for the selected questions. You will find the following options:
- Enable interview weighting: this setting is used to enable weighting for a specific table.
- Show unweighted total: this setting shows the unweighted number of interviews for data integrity, quality and credibility purposes.
Note: only interviews that have a value in the InterviewWeight variable are taken into account to ensure only valid responses contribute to the analysis. - Apply to all: this setting allows you to apply weighting to all selected tables at once.
If you can’t achieve your analysis goals with default available tables in table wizard, learning how Dashboard Reports work gives you the flexibility to create DIY custom tables that you need.
Limitations
- The {{NumberOfResponses}}, {{NumberOfMissings}} and {{NumberOfNa}} placeholders can not be used.
Instead if you want to show the weighted and/or unweighted number of responses using a placeholder (for charts) you can use the GetCellValue placeholders to refer to the summary rows of the specified table.
- Weighting can not be used for means, sums and standard deviations.
- Member variables can not be used, question and custom variables can be used instead.
- When you manually create tables with percentage distributions, ensure that you do not work with a custom formula ending with .ResponseCount, but with .Value. The .ResponseCount will always return the normal count, not the weighted count. When you use the table wizard for weighted tables this is automatically correctly executed:
