[R] cpower and censoring

[Frank E Harrell Jr]

Daniel Brewer wrote:
> I would like to do some power estimations for a log-rank two sample test
> and cpower seems to fit the bill.  I am getting confused though by the
> man page and what the arguments actually mean. I am also not sure
> whether cpower takes into account censoring or not.
> 
> Could anyone provide a simple example of how I would get the power for a
> set control/non-control clinical trial where censoring occurs at an
> estimated rate with an estimated drop out rate.
> 
> Quite confused about this.
> 

cpower handles censoring by specification of an accrual and follow-up 
periods and the event probability by reference time tref.  I believe the 
censoring distribution is assumed to be uniform, i.e., that accrual 
occurs at a uniform rate.

To have more control over the situation, the simulation-based spower 
function is one to look into.

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