Date: Fri, 26 Mar 2010 01:53:43 -0400
Reply-To: Francois van der Walt <francoisw@GJI.COM.AU>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Francois van der Walt <francoisw@GJI.COM.AU>
Subject: Re: Which Statistical test to use?
As someone that compares the success of marketing campaigns on a daily basis
your question triggered a mountain of possibilities.
Most of my results needs to be communicated to non statisticians so I
daresay I favor graphical techniques.
Here is some techniques I suggest you look at:
1) Box and Whisker plots (replaces the t-test)
2) Shewhart/XmR charts (Page 122 Making Sense of Data by Donald Wheeler)
3) Lift curve (http://www.information-management.com/news/5329-1.html gives
A simple but effective graph that provides a lot of insight is a timeplot of
sales per customer for the two groups.
If you want to provide some sample data I will be able to provide better
suggestions and potentially some code.
Francois van der Walt (GJI www.gji.com.au)
On Wed, 24 Mar 2010 16:08:00 -0400, SUBSCRIBE SAS-L Anonymous
>I currently measure success of Marketing campaigns. Success is based on a
>few different factors such as changing revenue and reducing churn.
>I use t-test to measure the difference in revenue across the whole
>targeted vs a control group. And use Chi squared test for measuring churn.
>I measure this seperately each month.
>I have a couple of questions.
>1. Are these the best tests to use?
>2. How can I compare month on month. I use 0.1 as being statistically
>significant but if a campaign is 0.2 three months in a row could I say
>that overall the campaign is significant or would I have to role up all
>the data together?
>Basically I'm looking to find a way of measuring 3 months worth of a
>campaign together but without having to gather all the data in SAS again.
>Any help appreciated - hopefully the above all makes sense.