[R] Programme Maxstat

[Torsten Hothorn]

On Tue, 12 Aug 2003, Ciril Rozman wrote:

> Sirs,
> I have recently been interested in your Maxstat. 

For questions related to add-on packages please at least cc to the
maintainer since some might not read r-help regulary.

> I have computed 
> with my own programme the ranks (by using the Kaplan-Meier method and
> the
> log-rank test) with the formula (Observed-Expected)/(SQR Var). 

The problem is what "SQR Var" really means. The maxstat package uses a
reformulation of the logrank test as a linear rank statistic with special
scores and, more important, uses a conditional variance estimator which is
slightly different from the one implemented in "survdiff", for example.

> The
> results are similar but not exact to the M value obtained with the
> Maxstat. I would like to know whether you are using some correction or
> adjustment in computing the different ranks. 

The details can be found in the references cited in the documentation of
the package, especially Hothorn & Lausen, 2003, CSDA, freely available
from

http://www.elseviermathematics.com/mathematicsweb/show/Main.htt?Name=&Source=Journal%2Ehtt&Key=Biostatistics&Results=Journal%2Ehtt&External=mathematicsweb%2F15

Best,

Torsten

> Thank you very much for
> your help.
> 
> Dr. C. Rozman
> Professor of Medicine, Emeritus
> University of Barcelona, Spain
> E-mail : rozman at medicina.ub.es
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-help
> 
>