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

[Dieter Menne]

Serguei Kaniovski <Serguei.Kaniovski <at> wifo.ac.at> writes:

> 
> 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")
> 
> SORT <- c("GER","BEL","DEN","AUT")
> 
> I need to compute the correlation between countries in the variable "x" 
> in such a way that the rows & columns of the resulting correlation 
> matrix are not in an alphabetical order but in the order of a given 
> factor vector - here SORT.
> 

This boils down to : how do I reorder a factor not to use alphabetical order?
There are several reorders around, and you could do it with base function in R,
but I find the following solution using package gdata the most readable:

library(gdata)
df <- data.frame(country = sample(c("AUT","BEL","DEN","GER"),10,TRUE))
str(df)

SORT <- c("GER","BEL","DEN","AUT")
df$countryS = reorder(df$country, new.order=SORT)

Dieter