How to Add Covariates to Analysis?
Covariates are variables that relate to the respondent that a researcher believes will have a substantial impact on their preferences for features. Including covariates in Hierarchical Bayesian (HB) models can help reduce variability or noise, and improve the estimates of the partworth utilities.
You can set up covariates by setting up segmentation for Generic Conjoint, Brand-Specific Conjoint, MaxDiff analysis, Brand-Price Trade-Off, as well as HB modelling-enabled Claims Test and Product Variant Selector reports.
To do this, simply navigate to the Segmentation tab, and click on to apply the segment as a covariate in your analysis.
Generation of covariate matrix
When covariates are added, a covariate matrix is generated and applied in the HB calculation with the following conditions:
- The first column is all zeros, and subsequent columns are zero-centered.
- Any redundant columns are skipped automatically.
- Small segments or highly correlated segments are ignored.
The complete covariate matrix is available in the Excel report’s data dump.
Example: Including respondent location as covariates
The following example illustrates the subgroup analysis of preferences and diagnostics for different product names, comparing USA and non-USA segments without any adjustments.
The next example used the same data set but included respondent locations (USA and non-USA) as covariates in the analysis.
In the above examples, including respondent locations as covariates in the analysis allows us to observe larger differences between segments.