Visualization by Covariate
An example of a covariate profile plot is given below.
Each line corresponds to a covariate -- each line is a covariate profile. Plotting these profiles together allows the user to visually compare how covariate distributions change as a function of the subgroup-specific treatment effect. If the covariate is an effect modifier, we would expect to see a nonlinear profile, or a profile with a slope different from zero. The user is also able to hover their mouse pointer over any point in the graph to see what covariate the datum corresponds to [#4 in future update].
When an extreme value has been identified, as well as the associated covariate, the user is able to use the Viz by Covariate visualization to isolate that covariate and look more closely at how the distribution of that covariate changes as a function of the subgroup-specific treatment effect.
The user is able to select which covariates appear in the plot by clicking the corresponding color on the right [show/hide all: #15 in future update]. The user can also further customize the plot by
- hiding/showing the connecting lines
- hiding/showing the plot points
After the user has identified specific covariates for further investigation, they can investigate using the "Viz by Covariate" feature. The user is able to select a covariate and view how it changes as a function of the subgroup-specific treatment effect, as well as compare subgroup-specific covariate means to the marginal mean.
