CODEWITHSUNDEEP
rmatrixtime-seriescorrelation
User42
User42

Reputation: 27

Correlate all pairs of values in a time series in r

I'm trying to correlate timeseries data within site pairs. I'm working with a data set that looks like this:

A= Site, B= Year and C= Abundance:

         A    B    C
 [1,]    1 2002   21
 [2,]    1 2004   25
 [3,]    1 2005   26
 [4,]    2 3003   24
 [5,]    2 2004   20
 [6,]    2 2005   20
 [7,]    3 2002   21
 [8,]    3 2003   22
 [9,]    3 2004   23
[10,]    3 2005   25

I want to split the data by column A to test correlation between each site pair (site 1 with 2 and 3, site 2 with 1 and 3 etc.)

I've tried: mydata.cor = cor(dat, method = c("spearman"))

But this just correlates the columns:

           A          B          C
A  1.0000000 -0.1287697 -0.1684834
B -0.1287697  1.0000000  0.4151682
C -0.1684834  0.4151682  1.0000000

Is there a way specify grouping values, in this case site category?

Upvotes: 1

Views: 254

Answers (1)

hi_everyone
hi_everyone

Reputation: 333

one way to do it would be to transform your data to wide format and then use cor() as you intended to (on the relevant columns).

Here's my reproducible code, I assumed that the 3003 in line 4 was a typo ...

library(data.table)

mat <- matrix(c(
    1, 2002,   21,
    1, 2004,   25,
    1, 2005,   26,
    2, 2003,   24,
    2, 2004,   20,
    2, 2005,   20,
    3, 2002,   21,
    3, 2003,   22,
    3, 2004,   23,
    3, 2005,   25),10,3,byrow = TRUE)

DT <- data.table(mat)
names(DT) <- c("A","B","C")

DT_wide <- dcast(DT, B~A)
cor(DT_wide[,-1], method="spearman", use="pairwise.complete.obs")

Upvotes: 2

Related Questions