When running a regression for driver analysis, each respondent is treated as one observation for the regression modeling. Many times you may ask the same question across many brands to the same respondent, and you don't care about what the driver is for an individual brand, but rather what drives the respondents' overall outcome. To do this, you need to treat the response for each respondent-brand pair as an observation rather than each respondent, which means you need to stack the data. This article describes how to perform driver analysis using data for multiple brands at the same time by stacking the data:
Requirements
- A driver analysis model, see How to Do Driver Analysis
- An Outcome variable set structured as a Numeric, or Ordered Categorical such as a likelihood to recommend variable set by brand.
- A Predictor(s) need to be a single Variable Set that has a Grid type structure such as a Binary - Grid or a Numeric - Grid, such as rating attributes by the same brands as in the outcome.
Method
- From the object inspector, select your driver analysis regression model
- From the object inspector, tick the box for Inputs > Stack data.
- Select the questions for Outcome and Predictors based on requirements above.
- Click Calculate.
Next
How to Select the Regression Type for Driver Analysis
How To Automatically Remove Outliers from Regression and GLMs