Date:         Thu, 21 Feb 2008 10:21:56 -0500
Reply-To:     susie.li@BOEHRINGER-INGELHEIM.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Susie C Y Li <susie.li@BOEHRINGER-INGELHEIM.COM>
Subject:      Re: question on suitablilty of cluster analysis
In-Reply-To:  A<10077203.1203547842896.JavaMail.root@mswamui-cedar.atl.sa.earthlink.net>
Content-Type: text/plain; charset="us-ascii"

I have once done a clustering analysis of customers based on each of their
monthly consumption volumes of a product over a two-years' period.  The
result of this clustering grouped these customers into "consumption trend"
groups (i.e., growth / decline / up-and-downer / flatliner groups).

Susie C Li


>The situation:  I have total catch data for fish collections (summed over
>species) for some 16 years. Each year we were supposed to take three
>samples at each of 4 sites and using two different gear types. I.e., we
>have
>
>16 years * 2 gears * 4 sites * 3 seasons which, by my cyphering, means 384
>samples. Of course, the world does not work like we would like it and
>sometimes a whole set of 8 samples will be missed if the water is too high
>and we can't go out and, on other times, a piece of gear may break during a
>survey and we can't reschedule a follow up.
>
>Note that I do not have reps so I have total catches, not means over reps.
>
>Would these data be suitable for a cluster analysis so that my boss can
>look at a plot and see something like years forming groups?
>
>Of course, there are  uncontrolled variables such as percent of aquatic
>weeds in which fish can hide and which we have not measured but that's the
>life of fish biologists.
>
>Thoughts and comments, both positive and negative will be most welcome.
>
>