0
$\begingroup$

I am trying to test the impact of a sale initiative on store revenue. I have created 10 matched pairs of stores and then randomly distributed one store in each pair into a test or control group.

I then ran the sale for 21 days within the test group (keeping control constant) and measured relative lift in revenue compared to baseline.

My questions

1.) Should I be using a paired t-test based on the previous store matching

2.) Should I be measuring across each day of the test such as:

        Control Test
1/1/2017    4   3
1/2/2017    3   4
1/3/2017    1   3
1/4/2017    2   3
1/5/2017    2   4
1/6/2017    1   7

Or measuring across my paired markets such as:

               Test Control
Market Pair 1   10  5
Market Pair 2   8   3
Market Pair 3   5   1
Improve this question
$\endgroup$
2
  • 1
    $\begingroup$ Would not use t-test. Numbers small. Stores not really paired, assigning a pair does not adjust for relative sales that would occur in each store. $\endgroup$
    – Carl
    Commented Jan 4, 2018 at 23:00
  • 1
    $\begingroup$ what is number of test ? 24 or 23 ? please elaborate your market pair concept. And indicate your important hypotheses you intend to test. $\endgroup$
    – user10619
    Commented Jan 5, 2018 at 1:14

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.