Skip to content

Modality Test

7 messages · Amin W. Mugera, Roger Koenker, Mark Difford +1 more

Amin W. Mugera
# 
Dear Readers:

I have two issues in nonparametric statistical analysis that i need
help:

First, does R have a package that can implement the multimodality test,
e.g., the Silverman test, DIP test, MAP test or Runt test. I have seen
an earlier thread (sometime in 2003) where someone was trying to write
a code for the Silverman test of multimodality. Is there any other
tests that can enable me to know how many modes are in a distribution?

Second, i would like to test whether two distributions are equal. Does R
have a  package than can implement the Li (1996) test of the equality
of two distributions? Is there any other test i can use rather than the
Li test?

Thank you in advance for your help.

Amin Mugera
Graduate Student
AgEcon Dept. Kansas State University
Roger Koenker
# 
the diptest package, perhaps?


url:    www.econ.uiuc.edu/~roger            Roger Koenker
email    rkoenker at uiuc.edu            Department of Economics
vox:     217-333-4558                University of Illinois
fax:       217-244-6678                Champaign, IL 61820
On Sep 9, 2008, at 11:23 AM, Amin W. Mugera wrote:

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Mark Difford
    

    

      
      #
    

  


  

      
Hi Amin,

            

      
      
      
    

        
Jeremy Tantrum (a Ph.D. student of Werner Steutzle's, c. 2003/04) did some
work on this. There is some useful code on Steutzle's website:

http://www.stat.washington.edu/wxs/Stat593-s03/Code/jeremy-unimodality.R

I used it last year when I was trying to solve the problem of how best to
compare lots of density curves (age distributions of 3 spp. of tree
euphorbias from about very different 35 sites). In particular I had to
ensure that I wasn't creating spurious bimodality at a particular age range
when combining sites.

You might find it useful. Feel free to contact me off list if the code has
gone, as I think I still have it (somewhere).

Regards, Mark.

        
Amin W. Mugera wrote:

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Mark Difford
    

    

      
      #
    

  


  

      
Whoops! I think that should be Stuetzle --- though I very much doubt that he
reads the list.

        
Mark Difford wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Mark Difford
    

    

      
      #
    

  


  

      
Hi Amin,

And I have just remembered that there is a function called curveRep in Frank
Harrell's Hmisc package that might be useful, even if not quite in the
channel of your enquiry. curveRep was added to the package after my
struggles, so I never used it and so don't know how well it performs (quite
well, I would think).

Regards, Mark.

        
Amin W. Mugera wrote:

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Bert Gunter
    

    

      
      #
    

  


  

      
Here is a function that tests for equality of however many distributions as
you like:

equaldist <- function(...){
## ... numeric sample vectors from the possibly different distributions to
be tested
## returns TRUE only if the distributions are the same
FALSE
}

;-)

-- Bert Gunter
Genentech 

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On
Behalf Of Mark Difford
Sent: Tuesday, September 09, 2008 1:23 PM
To: r-help at r-project.org
Subject: Re: [R] Modality Test


Hi Amin,

And I have just remembered that there is a function called curveRep in Frank
Harrell's Hmisc package that might be useful, even if not quite in the
channel of your enquiry. curveRep was added to the package after my
struggles, so I never used it and so don't know how well it performs (quite
well, I would think).

Regards, Mark.

        
Amin W. Mugera wrote:

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Mark Difford
    

    

      
      #
    

  


  

      
Hi Bert,

            

      
      
      
    

        
Thanks for the elegant code. Problem is, the result is elusive, constantly
slipping from view and then rolling into the dip. Perhaps it's broken in
this version of R (2.7.2.Patched). A fix would therefore be much
appreciated, as I may need to use it again in the near future. A little
example follows:

##
p1 <- rnorm(100, 10, 5)
p2 <- rnorm(100, 35, 5)
equaldist(p1, p2)
[1] FALSE
equaldist(c(p1,p2), c(p2,p1))
[1] FALSE

plot(density(c(p1, p2)), main="Twin Peaks")
lines(c(p1,p2))
?

Regards, Mark.

        
Bert Gunter wrote: