Re: st: RE: Cox regression for grouped survival data

[Steve Samuels <sjsamuels@gmail.com>]

Enzo, you can be as semi-parametric as you'd like with -cloglog-. With
enough data you can fit a separate integrated baseline hazard to each
interval.

But a parsimonious approach is to fit a flexible polynomial to the
period terms, perhaps a fractional polynomial (-fp-). In a similar
spirit, the packages -stpm- (Royston) and -stpm2- (Lambert) at SSC fit
restricted cubic splines to the survival function.

The shortcomings of the the standard Kaplan-Meier and Breslow estimates
(Breslow, 1972) were presented by Lambert and Royston at the 2009 UK
Stata meetings. Cox himself preferred a parametric approach (Reid, 1994,
p. 450), as did Efron (1988).

References:

Breslow, NE. 1972. Contribution to the discussion of the paper by DR
Cox. Journal of the Royal Statistical Society, Series B 34, no. 2:
216-217.

Efron, Bradley. 1988. Logistic Regression, Survival Analysis, and the
Kaplan-Meier Curve. Journal of the American Statistical Association
Journal of the American Statistical Association 83, no. 402: 414-425.

Lambert, P. C., & Royston, P. (2009). Flexible parametric alternatives
to the Cox model. UK Stata User Group.
http://www.stata.com/meeting/uk09/uk09_lambert_royston.pdf

Reid, Nancy. 1994. A conversation with Sir David Cox. Statistical
Science 9, no. 3: 439-455.

Steve

> On Aug 8, 2013, at 11:11 AM, enzo.coviello@tin.it wrote:
> 
> Hi Pradip,
> 
> thanks for further reference.
> 
> I believe that the approach suggested by Steve may be applied to this
> situation (the work of prof. Jenkins in this field is precious), but in
> strict sense it is a parametric approach.
> 
> The approach followed by Feuer and al. should be semiparametric, that
> is a Cox-like model.
> 
> Enzo
> 
> ----Messaggio originale----
> 
> Da: Pradip.Muhuri@samhsa.hhs.gov
> 
> Data: 8-ago-2013 15.23
> 
> A: "statalist@hsphsun2.harvard.edu"<statalist@hsphsun2.harvard.edu>
> 
> Ogg: st: RE: Cox regression for grouped survival data
> 
> Hello,
> 
> The following article has also used Cox regression for grouped survival
> analysis of the NHIS-Linked Mortality Public Use Data Files.
> 
> "21st-Century Hazards of Smoking and Benefits of Cessation in the
> United Stateshttp://www.nejm.org/doi/full/10.1056/NEJMsa1211128#t=articleMethods
> 
> Recently, Steve Samuels (StataList) has suggested to me that I use
> complementary log log models (-cloglog-) for my analysis of the same
> data (mentioned above) but for a different topic.
> 
> The following is from Steve, " see the Lesson 6 link to discrete data
> analysis on Stephen Jenkins's fine web page "Survival analysis with
> Stata""
> 
> (http://www.iser.essex.ac.uk/survival-analysis).
> --
> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-
> statalist@hsphsun2.harvard.edu] On Behalf Of enzo.coviello@tin.it
> 
> Sent: Thursday, August 08, 2013 3:26 AM
> 
> To: statalist@hsphsun2.harvard.edu
> 
> Subject: st: Cox regression for grouped survival data
> 
> Dear Stata Users,
> 
> recently,for developping the Cancer Surveillance Query System, Feuer
> and coll. of the NCI used a Cox Regression for Grouped survival data
> (Cancer. 2012 Nov 15;118(22):5652-62. doi: 10.1002/cncr.27615. Epub
> 2012 May 8).
> 
> Cox model can be sensitive to tied survival times and often we know
> only the number of events occurring within time intervals.
> 
> Often we can read that Cox model is really suitable for data where time
> is continuous.
> 
> I would ask further addresses on this topic and mainly if exists a
> 
> Stata command-resource to fit  a Cox regression for grouped survival
> 
> data.
> 
> Many thanks.
> 
> Enzo
> 
> ^^^^^^^^^^^^^^^^


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