Date:         Thu, 22 May 1997 07:35:53 -0500
Reply-To:     "Scheltema, Karen" <Karen.Scheltema@STATE.MN.US>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         "Scheltema, Karen" <Karen.Scheltema@STATE.MN.US>
Subject:      Re: breakpoints

I would also add that categorizing continuous data can be an effective way
of dealing with grossly non-normal distributions.
 
Karen Scheltema, MA MS
Statistician
State of Minnesota
Department of Human Services
Division of Performance Measurement and Quality Improvement
444 Lafayette Road
St. Paul, MN 55155-3865
(612) 296-4339         fax (612)215-5754
karen.scheltema@state.mn.us
 
 ----------
From: R. Allan Reese
To: SPSSX(r) Discussion
Subject: Re: breakpoints
Date: Thursday, May 22, 1997 6:26AM
 
On Wed, 21 May 1997, William B. Ware wrote:
 
> I have to ask why you want to "categorize" a continuous variable.  So much
> information is lost...
 
Nothing has to be lost.  The original values remain in the data file,
and probably in the active data since the procedure is to create a set
of dummies that cover the range of interest.  Why do it?  To get round
some of the assumptions of parametric analysis.  For example, you can
replace linear regressions by the equivalent analysis of covariance and
fit piecewise models.  It promotes thinking about your data - go do it!
 
R. Allan Reese                        Email: r.a.reese@ucc.hull.ac.uk
Head of Applications                  Direct voice:   +44 1482 465296
Academic Services Computing           Voice messages: +44 1482 465685
Hull University, Hull HU6 7RX, UK.    Fax:            +44 1482 466441