Re: st: Cluster, new dissimilarity measures, and sequence analysis

[Jean-Benoit Hardouin <jean-benoit.hardouin@neuf.fr>]

I had this problem to create a process to construct clusters of 
dichotomous variables with measures of proximity equal to the average 
weighted correlations conditional to the scores of the individuals (the 
obtained clusters of variables must verify unidimensionality and local 
independence, two fundamental properties of the Item Response Theory 
(IRT)). This method is known as the HCA/CCPROX (Zhang and Stout 1999). 
My solution was to create my own Stata procedure to realize this method. 
The obtained module is named hcaccprox (ssc describe hcaccprox - 
information about this program on http://anaqol.free.fr). The problem of 
this procedure is that I have not work with the clusters tools of Stata, 
so it is not possible for example to construct a dendogram with "cluster 
dendogram"...). More, this procedure permit to cluster variables (and 
not the observations). But perhaps the code can help you...
I hope this helps

***************************************
Jean-Benoit Hardouin
Regional Health Observatory
Orléans - France
Email : jean-benoit.hardouin@neuf.fr
***************************************

Matissa Hollister a écrit:


> Im hoping someone can help me solve this problem,
> although I’m beginning to think that it’s hopeless. 
> Basically I’ve created my own special measure of
> dissimilarity that I want to use for clustering, but
> I’m finding that there is no way to get Stata to allow
> me to use this new dissimilarity measure.  Any ideas
> of ways to get around this problem would be greatly
> appreciated.
>
> Basically, I am using a procedure called Optimal
> Matching, an algorithm designed to create a measure of
> dissimilarity between two sequences of data.  I am
> using it to identify people who have similar career
> patterns.  I’ve created a do-file that accomplishes
> the most difficult and unusual part of Optimal
> Matching, which is creating the measure of
> dissimilarity between each pair of sequences.  I now
> want to run a clustering procedure to identify groups
> based upon this dissimilarity measure.
>
> I found a post in the listserv archives (dated
> November 18, 2002) where someone wanted to do
> something similar (she wanted to create a geographic
> distance measure).  From the response I gather the
> calling and running of the dissimilarity algorithms
> occurs within the built-in stata command _cluster and
> is done within C, which is certainly beyond my
> programming abilities.  I’ve contemplated several
> possibilities and would love help or advice on any of
> them:
>
> 1)find a different software program that will allow me
> to easily input a new dissimilarity measure into a
> cluster command (preferably not expensive)
>
> 2)a way to alter Stata’s cluster command to allow for
> this new dissimilarity measure
>
> 3)a way to get around this problem, e.g.:
>
>   A.use the ParseDist command within cluster.ado to
> somehow cause the built-in command to call up a
> different distance command
>
>   B.ways to enter the data so that a built-in Stata
> dissimilarity measure will result in the same pairwise
> distances (difficult because the pairwise
> dissimilarities make up a multi-dimensional space, the
> whole point is that they are difficult to summarize in
> a few variables)
>
> 4)	write my own clustering procedure
>
> Please!  Any help would be gratefully accepted.  I
> know that several other researchers have already used
> Optimal Matching with clustering, so my guess is that
> option #1 might be the most viable one, but I’m not
> sure where to look.
>
> Thanks,
>
> Matissa
>
>
>
> 	
> 		
> __________________________________
> Do you Yahoo!?
> Vote for the stars of Yahoo!'s next ad campaign!
> http://advision.webevents.yahoo.com/yahoo/votelifeengine/
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>
>  

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/