Date: Thu, 13 Sep 2007 19:34:36 -0000
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
Subject: Re: Help with Factor Analysis
Content-Type: text/plain; charset="us-ascii"
Of course, apologies!
Could I create dummy variables for each product (a column for each
product with a 1 or 0) and run factor analysis/principal components on
On Sep 13, 8:12 pm, Paige Miller <paige.mil...@kodak.com> wrote:
> andyma...@googlemail.com wrote:
> > Hi everyone,
> > Hope someone can help, I have a dataset which contains some customer
> > purchase data, I have
> > Date of purchase
> > Customer Number (unique)
> > Group (From a segmentation exercise)
> > Product (the product bought)
> > Qty (the number of products bought)
> > I have created a sample dataset as an example below.
> > data example;
> > format Date ddmmyy10.;
> > input CustNumber Group $ Date ddmmyy10. Product $ Qty;
> > datalines;
> > 1 A 01/01/2001 Prod1 1
> > 1 A 09/01/2001 Prod6 2
> > 1 A 08/02/2001 Prod3 1
> > 1 A 05/04/2001 Prod2 1
> > 1 A 07/04/2001 Prod8 3
> > 1 A 05/08/2001 Prod4 1
> > 1 A 03/09/2001 Prod5 1
> > 6 B 01/09/2002 Prod2 1
> > 6 B 03/09/2002 Prod5 1
> > ;
> > run;
> > I want to understand what the underlying groupings of products are by
> > Group. For example, customers in Group A purchase Products Prod1,
> > Prod4 and Prod3 (i.e. the groupings of the products, but not being
> > concerned about chronology, so Prod1 followed by Prod2 is the "same"
> > as Prod2 followed by Prod1 for my purposes).
> > This feels as though it is a factor analysis piece of work (principal
> > components?). Could someone point me the right direction to show me
> > how to go about producing the analysis in SAS?
> You can't do factor analysis or principal components on categorical
> To understand what is going on, I suggest you create simple frequency
> tables of the number of purchases of each product by group. If you
> need statistical tests to tell you if the frequencies differ from
> group to group, you would want to specify what hypotheses you wish to
> test (I believe you would fit a log-linear model) and then use
> contingency table testing methods in PROC CATMOD.
> Paige Miller
> paige\dot\miller \at\ kodak\dot\com- Hide quoted text -
> - Show quoted text -