interpretation of varImpact results

[DS-Rodrigues]

Hi Chris,

Can I ask one last question about interpretation of results? Let's assume a continuous outcome that is the average number of ice cream sold by a company per month in the last year. It ranges between 1 and 300. In this case, higher values are better. When looking at the highest ranked variable, which is significant and consistent, we find a treatment-specific mean for the high-risk level (which yields the highest outcome) of 0.2 and the treatment-specific mean for the low-risk level of 0.04. Impact = 0.16. These 0.2 and 0.04 are very small and do not make much sense in practice given that this is the highest ranked variable and the outcome can go up to 300. How can we say that this variable is the most important potentially if the treatment-specific mean for the high-risk level is 0.2? This seems to call into question the validity of the estimated high- and low-risk levels. Can you please advise?

Once again, thank you.