Rolling correlation between lists in R

Question

I have two lists, listA and listB which I would like to calculate the correlation between them using a rolling window of size 4.

listA A B C D 1) 1 3 4 6 2) 6 9 11 1 3) 1 3 4 5 4) 8 4 5 6 5) 9 9 4 6 6) 1 5 6 6 7) 9 3 6 4 8) 6 7 8 9 listB A B C D 1) 1 3 4 3 2) 6 9 5 7 3) 1 1 4 5 4) 7 1 5 6 5) 9 9 3 6 6) 1 5 6 6 7) 9 9 6 4 8) 5 6 4 9

Specifically I wish to calculate the cor(listA[1:4,1],listB[1:4,1]) using a rolling window of size 4 which means I'll have 5 values for the correlation between listA and listB for column A. Once this is done for column A I then would like to do the same for column B, C and D.

I have previously achieved this using 'rollapplyr' but the object I was working on was an XTS object and not two individual objects which for my simple little brain made it easier to comprehend.

'result <- rollapplyr(XTSobject, 4, cor, by.column = FALSE)'

Can anyone suggest a way to do this between two lists?

Can you make listA and listB reproducible please? Are they data frames? — Sotos
– Sotos, Commented Jun 7, 2017 at 7:40

Roman · Accepted Answer · 2017-06-07 08:01:55Z

Using the data from the answer of Lukasz Derylo you can also try:

set.seed(1234) listA<-data.frame(matrix(rnorm(32), c(8,4))) listB<-data.frame(matrix(rnorm(32), c(8,4))) do.call(cbind, Map(function(x, y) sapply(1:5, function(z) cor(x[z:(z+3)], y[z:(z+3)])), listA, listB)) X1 X2 X3 X4 [1,] -0.2193800 0.3119567 0.31504929 0.7475560 [2,] -0.2669925 -0.4210403 0.43135275 0.9510322 [3,] -0.5657515 -0.6421284 0.82403697 0.2723139 [4,] -0.3829411 -0.6975032 0.09799502 -0.1750354 [5,] -0.8950722 -0.2397563 0.49121024 -0.5240474

Łukasz Deryło · Accepted Answer · 2017-06-07 08:26:15Z

Let me randomly generate listA and listB and work on them:

listA<-data.frame(matrix(rnorm(32), c(8,4))) listB<-data.frame(matrix(rnorm(32), c(8,4)))

Then, prepare array for you results:

result<-array(NA, c(5,4)) dimnames(result)<-list(paste0(1:5, '-', 4:8), colnames(listA))

Loop-in-loop now:

for(i in 1:5){ for (j in 1:4){ result[i,j]<-cor(listA[i:(i+3),j], listB[i:(i+3),j]) } }

Here is what I got:

result X1 X2 X3 X4 1-4 -0.24299058 -0.6359179 -0.3303014 0.46628496 2-5 -0.05606969 -0.9332142 -0.5414745 0.08460162 3-6 0.21371864 -0.5366821 -0.5799187 0.19306623 4-7 0.50005529 -0.4661710 0.2146866 0.49636429 5-8 0.52684198 -0.3768619 0.3599111 -0.93804204

The same could be achieved without looping by using the apply family instead.

you can always add set.seed() in the beginning of the script, when treating with pseudorandom numbers (such as rnorm) in order to get reproducible results
Thanks Lukasz, Can I ask how can the same be achieved with an apply, is it a nested apply that's needed?
Look at Jimbou's answer. It contains nice do.call-sapply combo :)

Collectives™ on Stack Overflow

Rolling correlation between lists in R

2 Answers 2

Comments

3 Comments

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

3 Comments

Related