Say I would like to investigate whether a particular product feature affects its sales. We would like to match products based on feature vs non-feature. But the sales also depends on the retailers who sell the products, like marketing expense, ect.

In this case, I would like to match at both product level and retailer level. Is there any matching method addressing the issue? Or just combine all the features, both products and retailers, together and do the matching as usual?

  • $\begingroup$ Please be more specific. It would be good if you write down your mathematical model. If $T$ is the treatment variable (the particular product feature you mention), you don't match based on $T$ but on some other variables. Do you mean matching on other product-level features and retailer-level features? $\endgroup$ – chan1142 Jan 11 '19 at 1:51
  • $\begingroup$ @chan1142, yes, I would like to match on other product-level features and retailer-level features. After some investigation, it seems that multi-level PSM would work. But do you have any other thought? Sorry for the confusion. $\endgroup$ – Tracy Yang Jan 11 '19 at 17:39
  • $\begingroup$ Thanks for clarifying it. Another strategy would be to match based on all variables with or without interactions between product-level characteristics and retailer-level characteristics. All would depend on the structure of your data set. $\endgroup$ – chan1142 Jan 14 '19 at 12:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.