Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: running a backward stepwise multivariate analysis on my specific dataset


From   Ronan Conroy <rconroy@rcsi.ie>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   Re: st: running a backward stepwise multivariate analysis on my specific dataset
Date   Wed, 3 Dec 2008 18:12:31 +0000

On 3 Dec 2008, at 16:32, Scott Gilmore wrote:

I have a dataset of N=50 patients and 20 variables. 13 are categorical and 7 are continuous variables. I have no missing values. My outcome is one of the variables and it is a binary outcome, 1= yes disease, 0= no disease.

Looks like a nightmare to me - I have had someone in my office with almost exactly the same ratio of patients to variables. The trouble is that you don't have enough data. And a stepwise model will shrink badly when applied to new data, so the clinical validity of the exercise is very doubtful.

I might recommend -mrgraph- for inspecting the binary variables - with the -tab- option you get a nice 'northen blot'

Try also clustering routines to see if you can make any sense of the predictors.

But avoid statistical significance tests for the moment. Your chances of false negative results are very high given the sample size, and stepwise methods will only confuse the issue by capitalising on unreproducible features of your data.


Ronan Conroy
=================================

rconroy@rcsi.ie
Royal College of Surgeons in Ireland
Epidemiology Department,
Beaux Lane House, Dublin 2, Ireland
+353 (0)1 402 2431
+353 (0)87 799 97 95
+353 (0)1 402 2764 (Fax - remember them?)
http://rcsi.academia.edu/RonanConroy

P    Before printing, think about the environment




*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index