Date: Mon, 8 Dec 2008 17:20:57 -0500
Reply-To: Gene Maguin <firstname.lastname@example.org>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Gene Maguin <email@example.com>
Subject: Re: mail quantities determination
Content-Type: text/plain; charset="us-ascii"
Yes, that's pretty much what I figured. Here is how I'm thinking of this
problem right now. Suppose your expected (i.e., predicted) response rate is
2%. Then the variance is .98*.02 = .0196 and the sampling variance, assuming
a sample of 10,000, is .0196/10000 = 1.96E-06 and the standard error is
.00138. If you take a 1SD confidence interval your response rate would be
.01862 to .02138, or 1862 to 2138 persons.
>>Let me try to explain it once more. We have a model which is used for
scoring and selecting records for mailing campaign. From the response rates
obtained from past mailing campaigns and several other factors, we normally
calculate the quantity to be mailed from the population. For example,
Population - 1 mil
Mail Quantity - 500K
After scoring the population with the model, the records are ranked from 1
-10 based on the scores (records with highest scores as 1, and records with
lowest scores as 10). To meet our required mail quantity of 500K, we would
have to select records from Decile 1-5.
But since we also want to find out how the model performs (in terms of
response rates) in the bottom deciles (i.e. deciles 6-10), we want to mail
small samples that fall in the deciles 6-10. So we select 10K records from
each of the bottom deciles (6-10). The 10K quantity from each decile is not
calculated using any scientific method. What I would like to know is if
there is any scientific method to determine the additional quantities for
just the bottom deciles.
To manage your subscription to SPSSX-L, send a message to
LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
For a list of commands to manage subscriptions, send the command