Re: st: Simplification of formula in logistic regression

[Mikkel Brabrand <mikkel@brabrand.net>]

All.

Thanks for all your input and suggestions! I agree that it is not optimal to loose information this way, but although computers are everywhere, it is not easy to have the algorithms installed in the software we use! I am not happy about this, but it is the pragmatic way I have to seek. 

Anyway, of cause, I have also completed a logistic regression and get slightly better results (compared using area under the ROC curve) than when using the highly simplified model. However, I have also tried to use cubic splines. But I need to recalculate the cubic splines in my validation cohort and have no idea on how to do this. Do you have any suggestions on how I can calculate the cubic splines in my new cohort using the knots and weights I found in my original, development, cohort?

Thanks!

Mikkel

Den 16/05/2011 kl. 20.43 skrev David Radwin:

> This is true, and it raises the issue of false precision, which is also to
> be avoided. An old joke illustrates the problem:
> 
> -------
> 
> A visitor to a natural history museum is looking at a fossilized dinosaur
> skeleton. A guard nearby tells the visitor, "That dinosaur died 90,000,006
> years ago."
> 
> "Really?" asks the visitor skeptically. "How did you arrive at that exact
> number?"
> 
> "It's simple," replies the guard. "It died 90,000,000 years before I
> started working here, and that was six years ago."
> 
> --------
> 
> David
> --
> David Radwin
> Research Associate
> MPR Associates, Inc.
> 2150 Shattuck Ave., Suite 800
> Berkeley, CA 94704
> Phone: 510-849-4942
> Fax: 510-849-0794
> 
> www.mprinc.com
> 
> 
>> -----Original Message-----
>> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-
>> statalist@hsphsun2.harvard.edu] On Behalf Of Marcello Pagano
>> Sent: Monday, May 16, 2011 11:22 AM
>> To: statalist@hsphsun2.harvard.edu
>> Subject: Re: st: Simplification of formula in logistic regression
>> 
>> Before knocking this request too much further, one should consider the
>> accuracy of the variables going into the equation.  Something like blood
>> pressure, which can be measured very accurately at any instant, can vary
>> tremendously a minute later.  One should not be fooled by apparent
>> accuracy of clinical measures.  The grandaddy (or grandmom??) of all
>> these is the Apgar score.  She wanted a measure of the babies at birth
>> based on what we would consider very, very loose measures --- e.g.
>> Reflex irritability (response of skin simulation to feet) : No response
>> (score of 0); Some motion (score of 1); and Cry (score of 2); or Color:
>> Blue:Pale (score of 0);  Body Pink: extremities blue (score of 1); and
>> completely pink (score of 2) --- and yet the use of this score has
>> proven to be a great advance in pediatrics.  An excellent read:
>> "The Score" by Atul Gawande
>> http://www.newyorker.com/archive/2006/10/09/061009fa_fact
>> 
>> m.p.
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/