R help - May 2011 - Confidence interval for difference in Harrell's c statistics (or equivalently Somers' D statistics)

Dear All,

I am trying to calculate a 95% confidence interval for the difference in two
c statistics (or equivalently D statistics).  In Stata I gather that this
can be done using the lincom command.  Is there anything similar in R?

As you can see below I have two datasets (that are actually two independent
subsets of the same data) and the respective c statistics for the variables
in both cases.  What I would now like to do is to prove that there is no
statistically significant difference between the statistics (between the dev
and val datasets.)

Any help would be much appreciated.
> rdev <- rcorrcens(Surv(stimes1,eind1)~gendat1+neurodat1)
> rdev
Somers' Rank Correlation for Censored Data    Response
variable:Surv(stimes1, eind1)

              C    Dxy  aDxy    SD    Z      P    n
gendat1   0.534  0.069 0.069 0.017 3.98 0.0001 1500
neurodat1 0.482 -0.036 0.036 0.011 3.18 0.0015 1500
> rval <- rcorrcens(Surv(stimes2,eind2)~gendat2+neurodat2)
> rval
Somers' Rank Correlation for Censored Data    Response
variable:Surv(stimes2, eind2)

              C    Dxy  aDxy    SD    Z     P    n
gendat2   0.543  0.085 0.085 0.017 4.94 0e+00 1500
neurodat2 0.481 -0.038 0.038 0.011 3.44 6e-04 1500

Many thanks,
Laura
P.S. I'm using Windows XP, R 2.9.2

	[[alternative HTML version deleted]]

On May 5, 2011, at 8:20 AM, Laura Bonnett wrote:
> Dear All,
>
> I am trying to calculate a 95% confidence interval for the  
> difference in two
> c statistics (or equivalently D statistics).  In Stata I gather that  
> this
> can be done using the lincom command.  Is there anything similar in R?
Have you looked at rcorrp.cens {Hmisc}?

<snipped>
>
> 	[[alternative HTML version deleted]]

Please post in plain text.

-- 
David Winsemius, MD
West Hartford, CT