[R] Two-way non-parametric ANOVA?

Frank E Harrell Jr f.harrell at vanderbilt.edu
Wed Jan 23 23:45:38 CET 2008


David Airey wrote:
> While I defer to Frank as far as expertise goes, not having access to  
> his paper, I googled, and I just read that this kind of model is not  
> possible when there are a lack of ties for ranks in the data, and adds  
> an intercept for every rank in the data set. So while interpretation  
> doesn't get more difficult with additional ranks, unlike the  
> multinonimal model, Frank's suggestion would not fly for continuous  
> data that was transformed to ranks, as opposed to data collected as  
> ordered categories, where it would be preferable.

That is incorrect, except for the part about needing one intercept per 
unique value of Y, less one.  This only presents a RAM and execution 
time problem.  Sometimes I put continuous data into 100 quantile groups 
to make this fast.  SAS JMP (after I suggested this to them in 1982) 
handles an arbitrarily high number of unique Y values as they made use 
of a patterned covariance matrix, which I didn't in lrm.

So the PO model is a competitor of ordinary regression.  A new option in 
predict.lrm makes it easy to get the predicted mean in this situation.

Frank
> 
> On Jan 23, 2008, at 10:52 AM, Frank E Harrell Jr wrote:
> 
>> David Airey wrote:
>>>> We need a two-way non-parametric ANOVA in order to analysis  
>>>> properly  some ecological data, do you know any reference in R? or  
>>>> how to do  it? Thank you very much All the best diana
>>>>
>>> A couple more references here are below. I liked the Scheirer  
>>> reference.
>>> /*
>>> Scheirer CJ, Ray WS, Hare N (1976) The analysis of ranked data  
>>> derived
>>> from completely randomized factorial designs. Biometrics 32:429-434
>>> Groggel DJ, Skillings JH (1986) Distribution-free tests for main  
>>> effects
>>> in multifactor designs. American Statistician 40:99-102
>>> */
>> My guess is that those are a bit out of date, especially the first  
>> one, when compared to the PO model.
>> Cheers
>> Frank
>>
>>> ______________________________________________
>>> 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.
>>
>> -- 
>> Frank E Harrell Jr   Professor and Chair           School of Medicine
>>                    Department of Biostatistics   Vanderbilt University
> 
> ______________________________________________
> 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.
> 


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



More information about the R-help mailing list