Re: [R] How to map clusters to a correlation matrix

2007-11-20 Thread Serguei Kaniovski 


To Walter,

I am building a model of voting in the EU Council of Ministers. A voting
scenario assumes the probabilities of yes votes, possible different for
each country, and correlation coefficient between them.

The country variables (data) reflect the background characteristics that
are relevant for the likelihood of bloc formation in the Council. I would
like to cluster the countries in what I call probabilistic voting blocs,
and construct a correlation matrix between votes in different blocs and
between votes in each bloc.

I then use the probabilities and correlation coefficients to construct a
joint probability distribution on the set of all conceivable voting
outcomes and thus compute the probabilities of voting outcomes that are of
interest, such as probability of a country casting a decisive vote, etc.

Serguei

Austrian Institute of Economic Research (WIFO)

P.O.Box 91  Tel.: +43-1-7982601-231
1103 Vienna, AustriaFax: +43-1-7989386

Mail: [EMAIL PROTECTED]
http://www.wifo.ac.at/Serguei.Kaniovski
[[alternative HTML version deleted]]

__
R-help@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.

[R] How to map clusters to a correlation matrix

2007-11-20 Thread Serguei Kaniovski 

Dear All,

I have several socio-economic and geographic variables for the 27 EU
countries. I would to use these data to derive a correlation matrix between
groups of countries (for a different application).

I thought of using kmeans to cluster the groups, and then calibrate between
group correlations using distances between the centroids, and within group
correlations using distances in a cluster to the own centroid. To calibrate
is to transform a distance to a (positive) correlation coefficient using
some suitable function. Positive correlations reflect the strength of
common tendencies among the countries.

All the above seems crude to me, especially as you have to choice a
transformation function for a distance to a correlation coefficient. Are
there any better methods to do this?

Thanks is advance,
Serguei
[[alternative HTML version deleted]]

__
R-help@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.