BioProgram BioProgram - 2 months ago 8
R Question

Extracting and formatting results of cor.test on multiple pairs of columns

I am trying to generate a table output of a correlation matrix. Specifically, I am using a for loop in order to identify a correlation between all data in columns 4:40 to column 1. While the results of the table are decent, it does not identify what is being compared to what. In checking attributes of

cor.test
,I find that data.name is being given as
x[1]
and
y[1]
which is not good enough to trace back which columns is being compared to what. Here is my code:

input <- read.delim(file="InputData.txt", header=TRUE)
x<-input[,41, drop=FALSE]
y=input[,4:40]
corr.values <- vector("list", 37)
for (i in 1:length(y) ){
corr.values[[i]] <- cor.test(x[[1]], y[[i]], method="pearson")
}
lres <- sapply(corr.values, `[`, c("statistic","p.value","estimate","method", "data.name"))
lres<-t(lres)
write.table(lres, file="output.xls", sep="\t",row.names=TRUE)


The output file looks like this:

statistic p.value estimate method data.name
1 -2.030111981 0.042938137 -0.095687495 Pearson's product-moment correlation x[[1]] and y[[i]]
2 -2.795786248 0.005400938 -0.131239287 Pearson's product-moment correlation x[[1]] and y[[i]]
3 -2.099114632 0.036368337 -0.098908573 Pearson's product-moment correlation x[[1]] and y[[i]]
4 -1.920649487 0.055413178 -0.090571599 Pearson's product-moment correlation x[[1]] and y[[i]]
5 -1.981326962 0.048168291 -0.093408365 Pearson's product-moment correlation x[[1]] and y[[i]]
6 -2.80390736 0.00526909 -0.131613912 Pearson's product-moment correlation x[[1]] and y[[i]]
7 -1.265138839 0.206482153 -0.059798855 Pearson's product-moment correlation x[[1]] and y[[i]]
8 -2.861448156 0.004415411 -0.134266636 Pearson's product-moment correlation x[[1]] and y[[i]]
9 -2.103403363 0.035990039 -0.099108672 Pearson's product-moment correlation x[[1]] and y[[i]]
10 -3.610094985 0.000340807 -0.168498786 Pearson's product-moment correlation x[[1]] and y[[i]]


Clearly, this is not perfect as rows are numbered and can't tell which correlation is to what. Is there a way to fix this? I tried many solutions but none worked.I know that the trick must be in editing the
data.name
attribute however I couldn't figure out how to do that.

Answer

Here's a way to return a data frame with all the cor.test results that also includes the names of the variables for which each correlation was calculated: We create a function to extract the relevant results of cor.test then use mapply to apply the function to each pair of variables for which we want the correlations. mapply returns a list, so we use do.call(rbind, ...) to turn it into a data frame.

# Function to extract correlation coefficient and p-values
corrFunc <- function(var1, var2, data) {
  result = cor.test(data[,var1], data[,var2])
  data.frame(var1, var2, result[c("estimate","p.value","statistic","method")], 
             stringsAsFactors=FALSE)
}

## Pairs of variables for which we want correlations
vars = data.frame(v1=names(mtcars)[1], v2=names(mtcars)[-1])

# Apply corrFunc to all rows of vars
corrs = do.call(rbind, mapply(corrFunc, vars[,1], vars[,2], MoreArgs=list(data=mtcars), 
                              SIMPLIFY=FALSE))

     var1 var2   estimate      p.value statistic                               method
cor   mpg  cyl -0.8475514 9.380327e-10 -8.747152 Pearson's product-moment correlation
cor1  mpg disp -0.7761684 1.787835e-07 -6.742389 Pearson's product-moment correlation
cor2  mpg   hp  0.4186840 1.708199e-02  2.525213 Pearson's product-moment correlation
cor3  mpg drat  0.6811719 1.776240e-05  5.096042 Pearson's product-moment correlation
cor4  mpg   wt  0.4802848 5.400948e-03  2.999191 Pearson's product-moment correlation
cor5  mpg qsec  0.6640389 3.415937e-05  4.864385 Pearson's product-moment correlation
cor6  mpg   vs  0.5998324 2.850207e-04  4.106127 Pearson's product-moment correlation
cor7  mpg   am  1.0000000 0.000000e+00       Inf Pearson's product-moment correlation
cor8  mpg gear -0.8676594 1.293959e-10 -9.559044 Pearson's product-moment correlation
cor9  mpg carb -0.8521620 6.112687e-10 -8.919699 Pearson's product-moment correlation
Comments