Date:         Fri, 4 Mar 2005 11:51:46 -0500
Reply-To:     Wensui Liu <liuwensui@gmail.com>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Wensui Liu <liuwensui@GMAIL.COM>
Subject:      Re: knowledge discovery vs. hypothesis
Comments: To: "Li,Tom" <Li.Tom@endo.com>
In-Reply-To:  <4A6EB3AE96290A4CBC9337ABDBC847070FF07A@PA100-EX-01.Endo.com>
Content-Type: text/plain; charset=US-ASCII

Thank you so much, Tom. Your suggestion is indeed helpful.

But my problem is that before you see the data and investigate the
data structure, how do you know what kind of model or analysis you are
going to use?

It is like before meeting the girlfriend, you have to decide whether
to go out with her, no matter she's goodlooking or not.


On Fri, 4 Mar 2005 11:37:45 -0500, Li,Tom <Li.Tom@endo.com> wrote:
> The analysis methods should be pre-specified, not after the data are
> collected.  Most of the results from afterward introduced methods are
> not worth anything.
>
> -----Original Message-----
> From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
> Wensui Liu
> Sent: Friday, March 04, 2005 8:46 AM
> To: SAS-L@LISTSERV.UGA.EDU
> Subject: knowledge discovery vs. hypothesis
>
> Sometimes, the results I got with relatively new methods, such as GAM,
> are not consistent with my clients' hypothesis or presumption, which
> might be reluctant to believe my result.
>
> I am wondering how often you met this kind of situation and how you
> handle it.
>


--
WenSui Liu, MS MA
Senior Decision Support Analyst
Division of Health Policy and Clinical Effectiveness