[R] help Dxy and C-index calculation

[Frank Harrell]

Replace Design with rms (for general reasons not related to your question;  
Design is about to be obsolete).

Negate Dxy.  The linear predictor for the Cox model is relative log  
hazard.  Higher hazard means shorter survival time.  For other survival  
models the model is stated in terms of survival time.

Frank


On Fri, 19 Aug 2011 08:30:50 -0500, Chafika.MAZOUNI at igr.fr  
<Chafika.MAZOUNI at igr.fr> wrote:

> Dear professor,
>
> I am currently using Design package and the cph formula for assessing  
> multivariable analysis.
>
> I am tryng to get the C-index for my survival model based on Dxy  
> coefficient.
>
> I am confused since there is a negative value.
>
> Do I need to used the absolute Dxy ?
>
>   index.orig     training         test     optimism index.corrected  n
> Dxy   -0.341357727 -0.344002740 -0.341357727 -0.002645013     
> -0.338712715 40
> R2     0.084694141  0.095440176  0.079077594  0.016362582      
> 0.068331560 40
> Slope  1.000000000  1.000000000  0.897999711  0.102000289      
> 0.897999711 40
> D      0.033368588  0.038458457  0.030983429  0.007475027      
> 0.025893561 40
> U     -0.002890981 -0.002968495  0.003412045 -0.006380539      
> 0.003489558 40
> Q      0.036259569  0.041426951  0.027571385  0.013855566      
> 0.022404003 40
>
> Many thanks
>
> Dr MAZOUNI
> Institut Gustave Roussy
> Department of breast surgical oncology
> Villejuif, France


-- 
Frank E Harrell Jr Professor and Chairman      School of Medicine
                    Department of Biostatistics Vanderbilt University