R help - Sep 2010 - Sample size estimation for non-inferiority log-rank and Wilcoxon rank-sum tests

Hello Everyone,
 
I'm trying to conduct a couple of power analyses and was hoping someone
might be able to help. I want to estimate the sample size that would be
necessary to adequately power a couple of non-inferiority tests. The first would
be a log-rank test and the second would be a Wilcoxon rank-sum test. I want to
be able to determine the sample size that would be necessary to test for a 3 day
difference in median recovery time between 2 groups of cancer patients.
 
Both of these tests are infeasible using SAS Proc Power and I haven't been
able to find information about how to do them using either SAS or R.
 
Does anyone know how to perform either of these calculations? If so, I'd
greatly appreciate it if you could share a couple of examples.
 
 
Thanks,
 
Paul


	[[alternative HTML version deleted]]

I haven't done much with the type of data you're working with, but  
here is a post that lists a few packages for doing sample size  
calculations in R.  Perhaps one of them will be helpful.

https://stat.ethz.ch/pipermail/r-help/2008-February/154223.html

Andrew Miles

On Sep 27, 2010, at 2:09 PM, Paul Miller wrote:
> Hello Everyone,
>
> I'm trying to conduct a couple of power analyses and was hoping  
> someone might be able to help. I want to estimate the sample size  
> that would be necessary to adequately power a couple of non- 
> inferiority tests. The first would be a log-rank test and the second  
> would be a Wilcoxon rank-sum test. I want to be able to determine  
> the sample size that would be necessary to test for a 3 day  
> difference in median recovery time between 2 groups of cancer  
> patients.
>
> Both of these tests are infeasible using SAS Proc Power and I  
> haven't been able to find information about how to do them using  
> either SAS or R.
>
> Does anyone know how to perform either of these calculations? If so,  
> I'd greatly appreciate it if you could share a couple of examples.
>
>
> Thanks,
>
> Paul
>
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

	[[alternative HTML version deleted]]

On 09/27/2010 08:09 PM, Paul Miller wrote:
> Hello Everyone,
> 
> I'm trying to conduct a couple of power analyses and was hoping
> someone might be able to help. I want to estimate the sample size
> that would be necessary to adequately power a couple of
> non-inferiority tests. The first would be a log-rank test and the
> second would be a Wilcoxon rank-sum test. I want to be able to
> determine the sample size that would be necessary to test for a 3 day
> difference in median recovery time between 2 groups of cancer
> patients.
> 
> Both of these tests are infeasible using SAS Proc Power and I haven't
> been able to find information about how to do them using either SAS
> or R.
> 
> Does anyone know how to perform either of these calculations? If so,
> I'd greatly appreciate it if you could share a couple of examples.

I don't think it is possible in general. You need to put up some
assumptions about the shape of the distributions.

It is a generic problem with nonparametric tests that they are only
nonparametric under the null hypothesis (so to speak). I.e., the null
distribution of the test statistic does not depend on the distribution,
but as soon as you get to matters of efficiency and power, it begins to
matter what the distribution and the alternative hypothesis is.



-- 
Peter Dalgaard
Center for Statistics, Copenhagen Business School
Phone: (+45)38153501
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com