Weighting Basics


In an ideal dataset, your respondents are representative of the target population. In reality, this can be difficult to achieve, so judicious weighting can be used to achieve the desired proportions. Many datasets will contain a weight variable for this purpose. If your dataset does not contain a weighting variable, MarketSight's weighting feature enables you to create a new one. You can also use it to edit an uploaded weight variable within the MarketSight application.

For example, let's say that 70% of respondents in our survey are doctors, while only 30%of respondents are surgeons. We know that surgeons compose a larger percentage of the general population (40%), however, so we adjust by weighting the data accordingly.

Create a Weighting variable

  1. Click the Weighting button on the Variable View page
  2. Choose either "Create new Weight Variable" or "Update existing Weight Variable"
  3. If creating a new Weight Variable, specify the name and code, or
  4. If updating an existing Weight Variable, select the category and variable to update
  5. Select variables to weight from the left box and drag them to the box on the right
  6. Assign desired Target % for each value (must add to 100%)
  7. Click Create
  8. Click Revise if changes are necessary, or
  9. Click Done to close all weighting dialogs

Weighting Terminology Used in MarketSight

Term Definition
Raking MarketSight's default weighting method, also known as rim weighting. This approach is best used if you do not know the target values for all possible combinations. Accuracy is dependent on how close your sample matches your target population.
Tolerance Specifies how close the finished weighted counts for each value should be to the specified target counts.
Max Iterations Limits the maximum number of iterations run by the algorithm before stopping.
Effective Base Effective base = (sum of weight factors) squared divided by the sum of the squared weight. Using the effective base provides for accurate statistical results for weighted data.
Weighting Efficiency Effective base divided by sample size. Indicates how much statistical power is lost by weighting. The lower the number, the more power is lost.
Average Weight The sum of weights for all records divided by the number of records.  The smaller the number, the better.
Min/Max Weight The lowest and highest values generated for the weight variable.
Status Displays the number of iterations required to achieve the desired population proportions.
Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.