[R] Compute correlation matrix for panel data with specific ordering

[John Kane]

Have a look at str(df).  Those values are being interpreted as factors not numbers.  I don't think this is what you want.

--- On Mon, 6/29/09, Serguei Kaniovski <Serguei.Kaniovski at wifo.ac.at> wrote:

> From: Serguei Kaniovski <Serguei.Kaniovski at wifo.ac.at>
> Subject: Re: [R] Compute correlation matrix for panel data with specific ordering
> To: r-help at r-project.org
> Received: Monday, June 29, 2009, 7:20 AM
> I apologize for not being specific
> enough in my previous posting. Assume you have panel data in
> the form:
> 
> df <- data.frame( cbind( rep( c( "AUT" , "BEL" , "DEN" ,
> "GER" ) , 4) , cbind( rep( c( 1999 , 2000 , 2001 , 2002 ) ,
> 4 ) ), sample( 10 , 16 , replace=T) ) )
> names(df) <- c( "country" , "year" , "x" )
> 
> 1. I would like to compute the correlation matrix between
> countries based on the annual observations of the variable
> x. I tried the following:
> library( combinat )
> 
> temp <- split( df$x, df$year )
> apply( combn(4,2) , 2 , function(x) cor( temp[[1]] ,
> temp[[2]] ) )
> 
> This gives wrong answer. Why?
> 
> 2. The pairwise correlations computed as above should be in
> the order:
> 
> GER with BEL, GER with DEN, GER with AUT, BEL with DEN, BEL
> with AUT, DEN with AUT.
> 
> That is, the correctly sorted vector of factors is:
> 
> SORT <- c( "GER" , "BEL" , "DEN" , "AUT" ) not c( "AUT"
> , "BEL" , "DEN" , "GER" )
> 
> May be there is an altogether better way of achieving what
> I want?
> 
> Serge
> 
> ______________________________________________
> 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.
> 


      __________________________________________________________________
Make your browsing faster, safer, and easier with the new Internet Explorer® 8. Optimized f
lorer/