[R] How do we do correlation for big matrices?

William Dunlap wdunlap at tibco.com
Sat Dec 26 20:14:02 CET 2015


Since you only want the diagonal of the correlation matrix, the following
will probably
do the job using less memory.  The mapply versions works on the data.frames
you supplied, but will not work on matrices - be careful not to conflate
the two classes of data objects.

  > vapply(colnames(df1), function(i)cor(df1[,i],df2[,i]), 0)
         site1        site2        site3        site4        site5
 site6        site7
  -0.540644946  0.006898188 -0.035279748 -0.261648270  0.274059055
-0.076396648   -0.147696334
         site8        site9       site10
  -0.138916728  0.330632540  0.366095090
  > mapply(FUN=cor, df1, df2)
         site1        site2        site3        site4        site5
 site6        site7
  -0.540644946  0.006898188 -0.035279748 -0.261648270  0.274059055
-0.076396648   -0.147696334
         site8        site9       site10
  -0.138916728  0.330632540  0.366095090
Compare to your:
  > diag(cor(df1,df2))
         site1        site2        site3        site4        site5
 site6        site7
  -0.540644946  0.006898188 -0.035279748 -0.261648270  0.274059055
-0.076396648   -0.147696334
         site8        site9       site10
  -0.138916728  0.330632540  0.366095090


Bill Dunlap
TIBCO Software
wdunlap tibco.com

On Sat, Dec 26, 2015 at 10:55 AM, Marna Wagley <marna.wagley at gmail.com>
wrote:

> Hi R users,
> I have a very big two matrices of 12 columns and over 0.5 million columns
> (50,4710) and trying to get correlation value between two tables but I
> could not compute it because of big files.
> Would you give me any suggestion on how I can do the correlations for the
> big files?
>
> I used the following codes and the example data.
>
> df1<-structure(list(X = structure(c(1L, 5L, 6L, 7L, 8L, 9L, 10L, 11L,
> 12L, 2L, 3L, 4L), .Label = c("env1", "env10", "env11", "env12",
> "env2", "env3", "env4", "env5", "env6", "env7", "env8", "env9"
> ), class = "factor"), site1 = c(0.38, 0.83, 0.53, 0.48, 0.66,
> 0.09, 0.21, 0.02, 0.76, 0.62, 0.2, 0.47), site2 = c(0.19, 0.14,
> 0.66, 0.35, 0.18, 0.24, 0.18, 0.2, 0.86, 0.06, 0.51, 0.29), site3 = c(0.95,
> 0.51, 0.91, 0.48, 0.74, 0.67, 0.34, 0.72, 0.43, 0.49, 0.1, 0.48
> ), site4 = c(0.89, 0.54, 0.93, 0.18, 0.99, 0.21, 0.69, 0.29,
> 0.89, 0.84, 0.45, 0.2), site5 = c(0.38, 0.37, 0.01, 0.26, 0.97,
> 0.49, 0.39, 0.31, 0.14, 0.83, 0.99, 0.2), site6 = c(0.68, 0.67,
> 0.6, 0.92, 0.01, 0.04, 0.49, 0.38, 0.5, 0.37, 0.51, 0.17), site7 = c(0.08,
> 0.54, 0.31, 0.3, 0.77, 0.39, 0.03, 0.51, 0.28, 0.32, 0.86, 0.95
> ), site8 = c(0.54, 0.26, 0.87, 0.91, 0.12, 0.51, 0.31, 0.67,
> 0.69, 0.79, 0.76, 0.08), site9 = c(0.1, 0.68, 0.17, 0.44, 0.78,
> 0.9, 0.16, 0.31, 0.13, 0.34, 0.9, 0.16), site10 = c(0.53, 0.31,
> 0.88, 0.61, 0.92, 0.44, 0.92, 0.94, 0.55, 0.8, 0.27, 0.07)), .Names =
> c("X",
> "site1", "site2", "site3", "site4", "site5", "site6", "site7",
> "site8", "site9", "site10"), class = "data.frame", row.names = c(NA,
> -12L))
> df1<-df1[-1]
>
> df2<-structure(list(X = structure(c(1L, 5L, 6L, 7L, 8L, 9L, 10L, 11L,
> 12L, 2L, 3L, 4L), .Label = c("env1", "env10", "env11", "env12",
> "env2", "env3", "env4", "env5", "env6", "env7", "env8", "env9"
> ), class = "factor"), site1 = c(0.36, 0.29, 0.09, 0.07, 0.82,
> 0.88, 0.59, 0.57, 0.2, 0.29, 0.76, 0.2), site2 = c(0.91, 0.87,
> 0.91, 0.54, 0.53, 0.2, 0.23, 0.16, 0.42, 0.44, 0.01, 0.29), site3 = c(0.96,
> 0.56, 0.34, 0.34, 0.6, 0.63, 0.28, 0.25, 0.73, 0.45, 0.88, 0.39
> ), site4 = c(0.73, 0.79, 0.39, 0.59, 0.63, 0.24, 0.69, 0.94,
> 0.07, 0.23, 0.01, 0.99), site5 = c(0.88, 0.18, 0.37, 0.24, 0.61,
> 0.61, 0.54, 0.71, 0.12, 0.82, 0.26, 0.5), site6 = c(0.43, 0.52,
> 0.01, 0.76, 0.41, 0.57, 0.08, 0.75, 0.82, 0.98, 0.61, 0.74),
>     site7 = c(0.84, 0.14, 0.96, 0.04, 0.41, 0.84, 0.26, 0.59,
>     0.29, 0.3, 0.76, 0.05), site8 = c(0.12, 0.18, 0.75, 0.23,
>     0.96, 0.64, 0.33, 0.61, 0.25, 0.13, 0.99, 0.6), site9 = c(0.26,
>     0.58, 0.32, 0.67, 0.11, 0.8, 0.87, 0.05, 0.03, 0.47, 0.95,
>     0.81), site10 = c(0.94, 0.63, 0.64, 0.5, 0.94, 0.75, 0.44,
>     0.57, 0.19, 0.23, 0.08, 0.18)), .Names = c("X", "site1",
> "site2", "site3", "site4", "site5", "site6", "site7", "site8",
> "site9", "site10"), class = "data.frame", row.names = c(NA, -12L
> ))
> df2<-df2[-1]
> df2
> # here I put only 12 columns, but as I mentioned above I have more than 1/2
> million columns
> cor_site<-data.matrix(diag(cor(df1,df2)))
> It works fine for a small data but this big files did not work.
>
> Thanks for your suggestions.
> MW
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.
>

	[[alternative HTML version deleted]]



More information about the R-help mailing list