Frailty term in survival analysis

["Monti, Gustavo" <Gustavo.Monti@wur.nl>]

Dear S-Users:

 

I have some questions related with the interpretation of the outcomes of S-plus for Cox models with frailties.

 

I have fitted a model  from data pooled from several  research groups  consisting in variables reflecting attributes of the individuals and  also I was interest in assess if there is any  important effect due to differences between groups (protocols, facilities, quality of personnel etc.).

 

For that purpose I used the Cox model with frailty proposed in the book of Therneau & Grambsch. Also they show a way to calculate the Kendall's tau for a gamma or gaussian frailty.  The first question is:  this correlation of subjects within group (KT) can be interpreted as  % of homogeneity of individuals due to the research group effect, or is the % of homogeneity of individuals due to their individual attributes and the rest is due to the group effect?

 

The second question is about the interpretation of Rsquare. Is it the proportion of the total variance explained by group effect?

 

Finally; Is there  other better option?  I have interval-censored data.

I have a S-plus 2000 R3 on a PC under Windows.

 

Thanks in advance;

 

My best regards;

Gustavo Monti
Wageningen University and Research
Department of Animal Sciences
Quantitative Veterinary Epidemiology Group
Marijkeweg 40
P.O.Box 338
6700 AH Wageningen
The Netherlands

E-mail: Gustavo.Monti@WUR.nl
Fax:++ 31 (0) 317 48 5006
Phone: ++ 31 (0) 317 48 4222
Web site: http://www.zod.wau.nl/qve/index.html