Video Guide to weighting in A&R: https://youtu.be/31IlsMVCWBc
It is common to apply weights to data to make adjustments between the sample data and the population that you wish to model.
In the Walr Platform, any numeric question (Decimal or Integer) can be used as a weight.
This article covers the basics on how to create, edit, and apply weights. If you're looking for information on how to...
1) Import an existing weight set, please see the articles on importing data: https://support.walr.com/support/solutions/folders/101000299877
2) Use the Script Weight and/or Numeric Questions tabs in the weighting wizard, please see the Advanced User section.
There are two options for how the weight value will be handled:
- Population Weight/RIM weighting - The value of the variable will be used and the weighted total will be the sum of all of the weights (population).
- Sample Weight/Cell weighting - The weight values will be normalised so that the weighted total matched the number of records in the data set (sample size). The normalisation calculation used is (Weight Value / Total of Weight Values) X Number of records.
TABLE OF CONTENTS
Create a Weight Set
1. From within most tools (Topline, Cross Table, Chart, Dig, Q-Table, Data) click Actions then Weighting.
- Note that the drop down menu will vary between tools so the location of Weighting within the menu will change. This example uses the Cross Table menu.
2. From the Create new weight tab select the numeric question(s) you wish to use in the weight set and click Next.
There is not technically a limit on how many variables you can interlock, but entering more than four may crash A&R if you're asking it to create combinations across a lot of answer codes (I tried with six variables just to see what would happen and it crashed).
More importantly, you will need to enter the exact sample/population size you want to use, so too many variables may not produce reliable results. For example, you may be able to calculate reliable numbers for your sample set for a combination like "age x gender x region" from census data or other reliable sources, but trying to weight by "age x gender x region x income" may not be accurate if, for example, I don't know the percenatge of the population who are: 15-24-years-old, male, living in eastern Norway, and earning between 200,000 and 400,000 Norweigan krones annually. My advice is to keep it to two or three variables maximum and if you're unsure consult someone in Marketing Sciences or with a statistics degree.
3. Select the weight type - Cell or Rim weighting and click Next.
3a) Cell weighting - when you want to interlock weights, use cell weighting. Another use maybe when you've over or under sampled for a group (for example say you wanted n=220 males and you got n=235 males) you can correct this using cell weighting. In the below example, I've selected Gender (A1) and Age (A3).
3b) RIM weighting - good for when you don't require interlocked weights and/or when you don't know the exact sample sizes, but have rough population percentages. You can have as many RIMs as you want.
Whatever type of weighting you choose, you will be given the option to toggle a code on/off (default is on). This can be useful when you have a code like "prefer not to say" which is something you cannot reliably weight against.
Cell example:
RIM example:
4. Provide the required information (Weight Name, Description, population) and relative weight values.
- Name: Give the weight a name (no spaces or special characters)
- Weight Description: Not mandatory, BUT this is what appears in the tables, if you're trying out multiple weighting schemes this is useful so you know exactly what's been applied. You can enter whatever text you want in here.
- Population: the overall sample size of your study. In this case n=989.
- Auto correct: ticking this box corrects any imbalances in your data and it makes your weights as efficient as possible.
- Relative Value boxes:
- RIM Weighting: Enter a value without the percentage sign or a population figure. (Note you can enter decimals e.g. 50.5). In this example, the population of Norway is 51% Female and 49% Male. Whatever you enter, it should add to 100%. Alternatively, you can enter (n=) 504 Female and (n=) 485 Male.
- Cell Weighting: Enter the percentage or sample size you want to use for each group that you're weighting (e.g. n=120 Males aged 15-24). Whatever you enter, it should add to 100% if using percentages or to the total sample size of your study. (In this case n=989). The screen shot below won't show all eight entries, but will give you an idea of what you may enter in each box.
Cell example (using sample figures):
RIM example (using percentages):
Click Finish. You've now created a weight set! BUT it will still need to be applied. The next sections cover how to edit weights you've created or imported and how to apply a weight set.
Manage an Existing Weight
Once created or imported a weight set(s) you will find your weight set(s) under the Existing Weight tab.
From here you can Edit or delete the weight set if needed.
Clicking edit will open this screen where you can adjust the population numbers you've entered:
The Script Weight and Numeric Questions tabs are covered in the Advanced User section.
Apply a Weight Set
Weight sets are applied to your data by selecting appropriate option under the Options menu:
- The location of the option will vary between A&R functions. In the Cross Table tool it is found under the Filter and Bases category and in the Chart tool it is found under the Bases and Documentation category. You can always use the search bar to search for "weight" and it will bring it up.
Weight sets that you have created in this project can be found In the pull down menu for Weight set as shown below. Select the desired weight set to apply it to the data.
Options for Applying Weights to your Analysis
Applying weights to all analysis features within A&R: Weights can be applied across ALL analytical tabs in Analysis & Reporting by using the Apply weights button shown below. Create and select a weight set following the steps above, then click Apply weights to activate the same weight set on all other tabs within Analysis & Reporting.
- You cannot undo the Apply weights action once it is selected. If you need to remove the weights you must enter each tab and remove them manually.
This option is available in the following tabs:
- Topline
- Cross table
- Charts
- Dig
- Q Table
Alternatively, you can apply weights to a selected table(s) or chart(s). To do this, select the variable you want to weight and leave the others unselected:
Then click on options and select your weight set:
Then click on "Set options" and the data will update for only the variable you've selected. This is useful for testing out different weight sets or if you need to display weighted and unweighted data:
In this example, I've applied an interlocked Cell weight for Table 1 and three RIM weights for Table 2. Note that in the below, the Income variable (A4) was dragged across twice so that I could test out two different weighting schemes. This means I can export both as well if I want to look at them side-by-side to assess the impact.
Example using Income (A4) to test two weighting schemes:
Table 1:
Table 2:
