Date:         Thu, 12 Mar 2009 15:03:05 -0400
Reply-To:     Peter Flom <peterflomconsulting@mindspring.com>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Peter Flom <peterflomconsulting@MINDSPRING.COM>
Subject:      Re: Proper Test for Data
Comments: To: Bminer <b_miner@LIVE.COM>
Content-Type: text/plain; charset=UTF-8

Bminer <b_miner@LIVE.COM> wrote
>
>I have data for two groups: Group A and Group B. They are created as
>being as close to each other as possible except for the fact that
>Group A had an intervention.
>
>I have two pieces of data on each individual in each group:
>
>Sales in 3 months before intervention
>Sales in 3 months after intervention
>
>There are zeroes in the data sets.
>
>Should I subtract before and after and then compare this difference as
>an independent two group test (non parametrically because of the >=0
>nature of the data)
>
>Or is there a better way? Ultimately I want to know if the
>intervention worked and use the control (group B) to help validate.


Do you have 2 pieces of information (3 mos before, 3 after) or 6 (month by month)?  I'll guess 2, since that's what you said, but 6 would be nice

How many people in each group?

There are several things you could do:
1) Subtract "before" from after, then to a t-test.  The t-test is reasonably robust to violations of normality, especially with eqaul group sizes, unless the variances are very unequal, and, while your data have 0's, the differences (before from after) may be relatively normal.    This is simple and easy to understand.

2) Do a PROC GLM with "before" as a covariate ... this lets you look at whether, say, the intervention had different effects at different levels of "before".  That might be very useful to you.

3) Do a permutation test   This makes very few assumptions (I think only exchangeablility)

and probably a few other things that I am not thinking of at the moment :-)


Before *any* of this, I'd look at some descriptive statistics and some graphs .... parallel box plots or dot plots, of "before" and "after" (box plot is N is large, dot plot if it's small)





HTH

Peter

Peter L. Flom, PhD
Statistical Consultant
www DOT peterflomconsulting DOT com