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]]
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]]
My guess is that a mapply would take forever to run. I would split it up into smaller blocks - not too large so the calculation can fit into the RAM, and not too small to make the calculation tun too long. Say 500 columns per block, that way each correlation matrix takes up 500*500*8 bytes = 1.9 MB, so a even the full 1000 blocks would fit into a reasonably sized RAM (hopefully R will do a garbage collection from time to time anyway). At the risk of tooting my own horn, library(WGCNA) ## For allocateJobs n = ncol(df1) blocks = allocateJobs(n, 1000) # With 1000 blocks, roughly 500 columns per block... results.lst = lapply(blocks, function(index) diag(cor(df1[, index], df2[, index]))); result = unlist(results.lst) I haven't tested this code, but it shouldn't be too far from correct. On Sat, Dec 26, 2015 at 11:14 AM, William Dunlap via R-help <r-help at r-project.org> wrote:> 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]] > > ______________________________________________ > 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.