Date:         Fri, 26 Jan 2001 17:21:49 -0500
Reply-To:     Sigurd Hermansen <HERMANS1@WESTAT.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Sigurd Hermansen <HERMANS1@WESTAT.COM>
Subject:      Re: Discretizing continuous vars (was Proc GLM)
Comments: To: Dale McLerran <dmclerra@MY-DEJA.COM>
Content-Type: text/plain; charset="iso-8859-1"

An age group may also represent a specific cohort in a study. Think of it as
a fuzzy set.  An investigator may want to treat its effect as separate
dimension of a model.  For example, an age group more heavily exposed to a
disease (say, direct or indirect exposure to Gulf War pathogens).   It seems
to me that it makes sense to try to measure real effects first, then fit
curves to the data that measure them.

As another example, we can think of time as a continuous variable, and we
might want to model the probability of a posting to the list as a function
of time.  We suspect that our friend Bill Viergever only works on Fridays.
Sure enough, discretizing time into Friday's and vineyard tour days gives us
a great model for the probability that Bill will post a message to SAS-L.
Good to hear from you again today, Bill.  It must be Friday.  Sig

<-----Original Message-----
<From: Dale McLerran [mailto:dmclerra@MY-DEJA.COM]
<Sent: Friday, January 26, 2001 3:29 PM
<To: SAS-L@LISTSERV.UGA.EDU
<Subject: Discretizing continuous vars (was Proc GLM)

<it needs to be understood exactly what discretizing the continuous
<predictor variable actually is doing: it allows the user to fit
<a nonlinear curve to the data.