[R] No parametric methods

[Greg Snow]

For power studies you need to think about what the data will look like under the alternative hypothesis.  Is the data shifted over a certain amount? (the most common assumption), or scaled? Or both? Or a completely different shape? Etc.

My preferred method for power studies in this case is to use simulation:

1. decide what you data is likely to look like (based on previous data, assumptions, ...)
2. decide how you will analyze the data (possibly iterate between 1 and 2)
3. write a function that simulates data under the alternative hypothesis, then analyzes it (using decisions from 1 and 2) and returns the p-value or test statistic.  The function will often have a parameter for sample size and a parameter for the size of the difference (scale, etc.).
4. use the replicate function to run your function a bunch of times.
5. the proportion of times that the above gives significant results is an estimate of the power.

Hope this helps,  

-- 
Gregory (Greg) L. Snow Ph.D.
Statistical Data Center
Intermountain Healthcare
greg.snow at imail.org
801.408.8111


> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of Alon Ben-Ari
> Sent: Tuesday, September 22, 2009 9:35 AM
> To: r-help at r-project.org
> Subject: [R] No parametric methods
> 
> Hello I am interested  in finding out a method of power analysis
> (effect
> size and sample size calculation ) using R in non parametric methods?
> 
> I am running  R  2.8.1 running on linux open SUSE
> 
> Any libraries or documentation , I was not bale to google up any.
> 
> Thanks in Advance,
> 
> Ben-Ari Alon, MD
> University of Pittsburgh.
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at 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.