[R] write a function to do pairwise calculation

[arun]

Hi,
Even if you have 15 matrices, you should be able to do it in the same way.

For example in the case of 5 matrices:
set.seed(24)
A<- matrix(sample(1:50,20,replace=TRUE),ncol=4)
B<- matrix(sample(40:60,20,replace=TRUE),ncol=4)
C<- matrix(sample(1:60,20,replace=TRUE),ncol=4)
D<- matrix(sample(1:30,20,replace=TRUE),ncol=4)
E<- matrix(sample(30:50,20,replace=TRUE),ncol=4)

mat1<- combn(LETTERS[1:5],2) 

#combn(1:15,2) ###For 15 matrices.  Change `1:15` by the names of the matrices


library(energy)
 res<-sapply(split(mat1,col(mat1)),function(.dat) dcor(get(.dat[1]),get(.dat[2]),1.5))
names(res)<-apply(mat1,2,paste,collapse="")


res
#       AB        AC        AD        AE        BC        BD        BE        CD 
#0.9435313 0.8097978 0.6819835 0.8065327 0.7639587 0.8123220 0.7919720 0.6971948 
 #      CE        DE 
#0.7002440 0.8506617 


Also, I dcor() needs matrices that are compatible.


Make sure that the dimensions are compatible:
There was a typo in my previous reply. It should be

cor(A,B)

set.seed(24)
 A<- matrix(sample(1:50,20,replace=TRUE),ncol=4)
 B<- matrix(sample(40:60,30,replace=TRUE),ncol=6)
C<- matrix(sample(1:60,20,replace=TRUE),ncol=5)
D<- matrix(sample(1:30,20,replace=TRUE),ncol=2)
E<- matrix(sample(30:50,20,replace=TRUE),ncol=5)

cor(A,B) #works
 dcor(A,B)
#[1] 0.9651803


cor(A,C)
#Error in cor(A, C) : incompatible dimensions
dcor(A,C)
#Error in .dcov(x, y, index) : Sample sizes must agree
 cor(A,D)
#Error in cor(A, D) : incompatible dimensions


A.K.



________________________________
From: Amanda Li <amandali at uchicago.edu>
To: arun <smartpink111 at yahoo.com> 
Sent: Wednesday, June 19, 2013 1:11 AM
Subject: Re: [R] write a function to do pairwise calculation



Hello,

Thanks for your help! I am a novice in R, and may I ask how I am supposed to do it if it is actually 15 matrices instead of 3?

I gave up dcor because the matrices are too large. I was thinking of cor(A,B)*cor(A,B)/length(cor(A,B)) (the length varies because the matrices are of different dimensions). May I ask how am I supposed to write the function to make it work?

And also could you recommend a book that is suitable for R novice? Thank you so much for your help! I really appreciate it!

Best,
Amanda


2013/6/18 arun <smartpink111 at yahoo.com>

Hi,
>
>You didn't provide any information about the package.  I guess it is from "energy". 
>
>
>library(energy)
>x <- iris[1:50, 1:4]   #examples given in the package
>y <- iris[51:100, 1:4]
>z<- iris[101:150,1:4]
>vec1<-c("x","y","z")
>mat1<- combn(vec1,2)
>
>sapply(split(mat1,col(mat1)),function(.dat) dcor(get(.dat[1]),get(.dat[2]),1.5))
>#        1         2         3
>#0.1862890 0.2331567 0.2303689
>
>
>A.K.
>
>
>
>
>
>
>----- Original Message -----
>From: Amanda Li <amandali at uchicago.edu>
>To: r-help at r-project.org
>Cc:
>Sent: Monday, June 17, 2013 1:14 PM
>Subject: [R] write a function to do pairwise calculation
>
>Hello,
>
>I want to write a function to do pairwise calculation, but I don' know how
>to write it. Could anyone help?
>
>i.e. I have A (2*3), B(3*3), C(4*3) three matrices. I want to calculate
>distance correlation between each pair of matrices using code "dcor". How
>do I write a function so that I can get the result as a matrix integrating
>all the  pairwise results?
>
>Thanks in advance for your help!
>
>Best,
>Amanda
>
>    [[alternative HTML version deleted]]
>
>______________________________________________
>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.
>
>