[R] Binary outcome with non-absorbing outcome state

Renaud Lancelot renaud.lancelot at cirad.fr
Sat Jul 30 10:11:23 CEST 2005

Frank Funderburk a écrit :
> Singer & Willett (2003) also cover this ground.
> Singer, JD & Willett, JB (2003).  Applied longitudinal data analysis:
> Modeling change and event occurrence. New Yok:  Oxford University
> Press.
> -----Original Message----- From: Frank E Harrell Jr
> <f.harrell at vanderbilt.edu> Sent: Jul 29, 2005 9:25 AM To: John Sorkin
> <jsorkin at grecc.umaryland.edu> Cc: R-help at stat.math.ethz.ch Subject:
> Re: [R] Binary outcome with non-absorbing outcome state
> John Sorkin wrote:
>> I am trying to model data in which subjects are followed through
>> time to determine if they fall, or do not fall. Some of the
>> subjects fall once, some fall several times. Follow-up time varies
>> from subject to subject. I know how to model time to the first fall
>> (e.g. Cox Proportional Hazards, Kaplan-Meir analyses, etc.) but I
>> am not sure how I can model the data if I include the data for
>> those subjects who fall more than once. I would appreciate
>> suggestions about a models that I could use, how I would quantify
>> the follow-up time, how I account for the imbalance in the data
>> (some subjects would contribute one outcome measure, others 
>> multiple measures), etc.
>> Many thanks, John
> A great reference for this is
> @Book{the00mod, author =               {Therneau, Terry and Grambsch,
> Patricia}, title =                {Modeling Survival Data: Extending
> the Cox Model}, publisher =    {Springer-Verlag}, year =
> 2000, address =              {New York} }
> Frank
>> John Sorkin M.D., Ph.D. Chief, Biostatistics and Informatics 
>> Baltimore VA Medical Center GRECC and University of Maryland School
>> of Medicine Claude Pepper OAIC
>> University of Maryland School of Medicine Division of Gerontology 
>> Baltimore VA Medical Center 10 North Greene Street GRECC (BT/18/GR)
>>  Baltimore, MD 21201-1524
>> 410-605-7119 ---- NOTE NEW EMAIL ADDRESS: 
>> jsorkin at grecc.umaryland.edu

Another possibility is to discretize the time, group the observations by
covariate pattern and use a beta-binomial model (accounting for possible
overdispersion caused by the within-subject repeated events) with a
cloglog link. When time interval is short (e.g., 1 day), this is
equivalent to a Cox prpoprtional hazards model. See:

Prentice, R.L., Gloeckler, L.A., 1978. Regression analysis of grouped
survival data with application to breast cancer data. Biometrics, 34: 57-67.


Prentice, R.L., 1986. Binary regression using an extended beta-binomial
distribution, with discussion of correlation induced by covariate
measurement errors. J.A.S.A. 81, 321-327.

and subsequent papers.

Function betabin in package aod (among others) allows to fit such models.



Dr Renaud Lancelot, vétérinaire
Projet FSP régional épidémiologie vétérinaire
C/0 Ambassade de France - SCAC
BP 834 Antananarivo 101 - Madagascar

e-mail: renaud.lancelot at cirad.fr
tel.:   +261 32 40 165 53 (cell)
         +261 20 22 665 36 ext. 225 (work)
         +261 20 22 494 37 (home)

More information about the R-help mailing list