0

I have two sets of continuous variables, weekly sales per customer. One set contains the sales of customers in the control group, and the other set contains the sales of customers in the test group. There is a lot of historical data. The variance between the groups is not presumed to be equal.

The control group does not receive an email promotion, while the test group does. My dataset contains weekly sales per customer for many weeks before and after the email promotion was sent to the test group. I wish to test if the email promotion had a statistically significant impact on sales. I want to compare the before and after weekly customer sales of the test group, and compare them to the sales of the control group.

What kind of statistical test is most appropriate in this case, where I have two groups, one of which receives the email promotion? My initial thought was an independent sample t-test, but now I am questioning that intuition because there are not just two independent samples, there is a before and after treatment condition on one of the groups.

Thanks!

PugFanatic
  • 1
  • You can probably find an answer here: https://stats.stackexchange.com/questions/3466/best-practice-when-analysing-pre-post-treatment-control-designs – kjetil b halvorsen Jan 28 '20 at 19:49
  • If you can relate the sales from each customer from before the email to after the email, then you have a paired t-test. If it is just a bunch of customers prior and different set of customers after then it is an independent two sample test. – Dave2e Jan 28 '20 at 20:18
  • 1
    @kjetilbhalvorsen thank you, that's a great resource – PugFanatic Feb 04 '20 at 21:21
  • @Dave2e thank you, I understand the two instances you mentioned. I think my situation is probably best represented by the independent two sample test since I want to maintain the two separate groups, only one of which receives the email – PugFanatic Feb 04 '20 at 21:28

0 Answers0