Notice: On March 31, it was **announced** that Statalist is moving from an email list to a **forum**. The old list will shut down at the end of May, and its replacement, **statalist.org** is already up and running.

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

From |
Mikkel Brabrand <mikkel@brabrand.net> |

To |
"statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |

Subject |
Re: st: Goodness of fit |

Date |
Wed, 4 May 2011 13:23:17 +0200 |

Thanks for the reply. The thing is, that I am validating the score and thus not performing a new logistic regression (and then unable to use the postestimation commands). I have calculated a coefficient in the development cohort and am reusing this in the validation set. This is why I need to calculate goodness of fit by hand. Mikkel Den 03/05/2011 kl. 14.24 skrev Neil Shephard: > On Tue, May 3, 2011 at 1:02 PM, mikkelbrabrand <mikkel@brabrand.net> wrote: >> I have developed a risk scoring system that I'm trying to validate. It was >> originally developed in one cohort, and I am now validating it in a new, >> independent, cohort. However, I am unsure how to test for goodness-of-fit >> using Hosmer-Lemeshow's test. >> >> First question: When validating my score, is it not correct that I should >> use >> the observed mortality (my endpoint) from the development cohort to estimate >> the expected mortality in my validation cohort? >> >> Second question: When calculating the Chi-square between the observed and >> expected mortality, should my formula not be: (observed-expected)^2/ >> (expected*1-expected/N (in this stratum))? >> >> Third question: How many degrees of freedome should I use? When I read the >> book by Hosmer and Lemeshow (Applied logistic regression), in the section on >> validation in an external cohort, I understand that the number of groups >> equal the degrees of freedom, which in different from the ordinary method in >> which the degrees of freedom is groups-2. Is this correct? > > You might want to check out -man logistic postestimation##estatgof- > which implements the Hosmer-Lemeshow test (and permits you to specify > the number of quantiles, and hence degrees of freedom). > > If you didn't know this existed you could have searched for it within > Stata using -findit hosmer- and sifter through the results. > > Neil > > > -- > “Truth in science can be defined as the working hypothesis best suited > to open the way to the next better one.” - Konrad Lorenz > > Email - nshephard@gmail.com > Website - http://kimura.no-ip.org/ > Photos - http://www.flickr.com/photos/slackline/ > > * > * 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/

**References**:**st: Goodness of fit***From:*mikkelbrabrand <mikkel@brabrand.net>

**Re: st: Goodness of fit***From:*Neil Shephard <nshephard@gmail.com>

- Prev by Date:
**st: out-of-sample prediction does not work after -regress- with lags** - Next by Date:
**st: Reporting pooled p-values smaller than e-40 with the metan command** - Previous by thread:
**Re: st: Goodness of fit** - Next by thread:
**st: RE: Goodness of fit** - Index(es):