[R] Bimodal deconvolution

[Peter Dalgaard]

Seth Imhoff wrote:
> Hi Everyone-
>
> After searching through posts and my favorite R-help websites I'm still confused about a problem.  I have data which is bimodal in nature, but there is no clearly obvious separation between the two peaks.  In programs such as Origin, I can deconvolute the two distributions and have it generate a "best guess" as to what the two subpopulations are which make up my dataset.  Below is a link to a figure which represents what I would like to accomplish here (it is the best I can do so far).
>
> http://nucleus.msae.wisc.edu/example.html
>   
That's not a well defined problem without further assumptions, so if you 
want to do what Origin does, you need to find out what Origin does.... 
(This is not easily fathomed by Googling.)

The key concept is "finite mixture model" an there's a CRAN task view on 
the topic:

http://cran.r-project.org/web/views/Cluster.html

("Deconvolution" is a somewhat different technique which is based on the 
assumption that there is an underlying "signal" which is blurred by some 
smoothing kernel. That doesn't really seem to apply here, but you could 
use the search facilities on www.r-project.org to investigate.)

-- 
   O__  ---- Peter Dalgaard             Øster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark          Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)                  FAX: (+45) 35327907