Correlation surface in R

Question

This is my third attempt on explaining my problem. Hope I can get it right this time, it is difficult problem to explain.

I have 90x19 matrix where each of the 90 rows is a series of 19 measurements on a certain wavelength. The 90 wavelengths goes from 400 to 700. Then I have a length 19x1 vector.

I want to create a matrix where each cell shows the Pearson correlation coefficient between the sum of each combination of wavelengths (for all the 19 values of the matrix) and the 19x1 vector.

Plotting this surface looks exactly like this

Please let me know if you need more information or a better explanation. Really need some help here ! :)

Bests

The vector is the following:

v<-c(116, 100, 148, 132, 81, 136, 145, 116, 87, 126, 62, 124, 129, 
108, 127, 134, 142, 99, 132)

And the data-frame header is the following:

data<-structure(list(`1` = c(2, 2, 2, 2), `2` = c(1, 1, 1, 1), `3` = c(2, 
2, 2, 2), `4` = c(2, 2, 2, 3), `5` = c(13, 14, 14, 15), `6` = c(2, 
2, 2, 2), `7` = c(2, 2, 2, 2), `8` = c(0, 0, 0, 0), `9` = c(5, 
5, 5, 5), `10` = c(0, 0, 0, 0), `11` = c(114, 119, 122, 125), 
    `12` = c(8, 8, 8, 8), `13` = c(7, 7, 7, 8), `14` = c(10, 
    10, 10, 10), `15` = c(12, 12, 12, 12), `16` = c(6, 6, 6, 
    6), `17` = c(4, 4, 4, 4), `18` = c(17, 18, 18, 19), `19` = c(13, 
    14, 14, 14)), .Names = c("1", "2", "3", "4", "5", "6", "7", 
"8", "9", "10", "11", "12", "13", "14", "15", "16", "17", "18", 
"19"), row.names = c("402.37700182806", "405.70862540288", "409.04076825642", 
"412.37342090064"), class = "data.frame")

You should provide some sort of [reproducible example](http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) with sample data to make it easier to help you. — MrFlick, May 30 '16 at 23:06
You need to exemplify the process with a small matrix and vector. I predict this explanation will not succeed without an example. — IRTFM, May 30 '16 at 23:06
Hey guys thanks for the advices, I have pasted the vector and the header of the dataframe, basically i would like to plot every correlation value for each row sum. e.g. cor((row1+row2)&v) — Emiliano, May 31 '16 at 00:40

Robert · Accepted Answer · 2016-05-31T02:39:18.557

I simulated the data (without correlation), but see if this is what you want:

nr=90
data<-data.frame(matrix(runif(nr*19,2,15),nrow=nr,ncol=19))
row.names(data)<-round(seq(400,700, length.out = nrow(data)),4)

cm=matrix(NA,nrow(data),nrow(data))
for (i in 1:nrow(data))
  for(j in i:nrow(data))cm[j,i]<- cor(apply(data[c(j,i),],2,sum),v)

colnames(cm)<-row.names(cm)<-round(as.numeric(row.names(data)),0)

#Build the plot    
library(lattice)
library(reshape)

cmd<-cbind(var=as.numeric(row.names(data)),data.frame(cm))
cmd<-cbind(melt(cmd, id=c("var")),var2=rep(as.numeric(row.names(data)),each=nrow(data)))

levelplot(value~var*var2,data=cmd,  
          col.regions=colorRampPalette(c("blue","green", "red")),
          at=seq(-1,1,0.1),
          xlab="x", ylab="x", main="")

Correlation surface in R

1 Answers1