Hi!

I'm doing a direct mail piece of 100,000 units.  I expect my response
rate to be about 3%.  How do I determine how many of the potential
100K recipients should be held aside as a control (i.e., not mailed)
so as to determine if my uplift is statistically significant? (at
confidence intervals of both 90% and 95%)

This is actually only one of my test cells for a series of direct mail
campaigns.  So, I'd like to understand how it is derived in order to
determine the appropriate sizes for the other tests as well.

Thanks!

---

Request for Question Clarification by mathtalk-ga on 14 Jul 2003 14:26 PDT

Hi, justwondering=ga:

Do you have an estimate of the response rate among those who will not
receive the mailing?  The point is that is the expected response of 3%
among recipients is slightly better than among the control group, a
larger sample is required in order to establish a significant
difference.

So an efficient approach would take advantage of any such additional
information that you have.

regards, mathtalk-ga