Last data update: 2014.03.03

 R: Determine highly correlated variables
FindCorrR Documentation

Determine highly correlated variables

Description

This function searches through a correlation matrix and returns a vector of integers corresponding to columns to remove to reduce pair-wise correlations.

Usage

FindCorr(x, cutoff = .90, verbose = FALSE)

Arguments

x

A correlation matrix

cutoff

A numeric value for the pair-wise absolute correlation cutoff

verbose

A boolean for printing the details

Details

The absolute values of pair-wise correlations are considered. If two variables have a high correlation, the function looks at the mean absolute correlation of each variable and removes the variable with the largest mean absolute correlation.

There are several function in the subselect package (leaps, genetic, anneal) that can also be used to accomplish the same goal.

Value

A vector of indices denoting the columns to remove. If no correlations meet the criteria, numeric(0) is returned.

Author(s)

Original R code by Dong Li, modified by Max Kuhn

References

Max Kuhn. Contributions from Jed Wing, Steve Weston, Andre Williams, Chris Keefer, Allan Engelhardt, Tony Cooper, Zachary Mayer and the R Core Team (2014). caret: Classification and Regression Training. R package version 6.0-35. http://CRAN.R-project.org/package=caret

See Also

leaps, genetic, anneal

Examples

corrMatrix <- diag(rep(1, 5))
corrMatrix[2, 3] <- corrMatrix[3, 2] <- .7
corrMatrix[5, 3] <- corrMatrix[3, 5] <- -.7
corrMatrix[4, 1] <- corrMatrix[1, 4] <- -.67

corrDF <- expand.grid(row = 1:5, col = 1:5)
corrDF$correlation <- as.vector(corrMatrix)
PlotCorr(xtabs(correlation ~ ., corrDF), las=1, border="grey")

FindCorr(corrMatrix, cutoff = .65, verbose = TRUE)

FindCorr(corrMatrix, cutoff = .99, verbose = TRUE)

# d.pizza example
m <- cor(data.frame(lapply(d.pizza, as.numeric)), use="pairwise.complete.obs")
FindCorr(m, verbose = TRUE)
m[, FindCorr(m)]

Results


R version 3.3.1 (2016-06-21) -- "Bug in Your Hair"
Copyright (C) 2016 The R Foundation for Statistical Computing
Platform: x86_64-pc-linux-gnu (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> library(DescTools)
> png(filename="/home/ddbj/snapshot/RGM3/R_CC/result/DescTools/FindCorr.Rd_%03d_medium.png", width=480, height=480)
> ### Name: FindCorr
> ### Title: Determine highly correlated variables
> ### Aliases: FindCorr
> ### Keywords: manip
> 
> ### ** Examples
> 
> corrMatrix <- diag(rep(1, 5))
> corrMatrix[2, 3] <- corrMatrix[3, 2] <- .7
> corrMatrix[5, 3] <- corrMatrix[3, 5] <- -.7
> corrMatrix[4, 1] <- corrMatrix[1, 4] <- -.67
> 
> corrDF <- expand.grid(row = 1:5, col = 1:5)
> corrDF$correlation <- as.vector(corrMatrix)
> PlotCorr(xtabs(correlation ~ ., corrDF), las=1, border="grey")
> 
> FindCorr(corrMatrix, cutoff = .65, verbose = TRUE)
Considering row	 3 column	 2 value	 0.7 
  Flagging column	 3 
Considering row	 2 column	 5 value	 0 
Considering row	 2 column	 1 value	 0 
Considering row	 2 column	 4 value	 0 
Considering row	 5 column	 1 value	 0 
Considering row	 5 column	 4 value	 0 
Considering row	 1 column	 4 value	 0.67 
  Flagging column	 4 
[1] 3 4
> 
> FindCorr(corrMatrix, cutoff = .99, verbose = TRUE)
Considering row	 3 column	 2 value	 0.7 
Considering row	 3 column	 5 value	 0.7 
Considering row	 3 column	 1 value	 0 
Considering row	 3 column	 4 value	 0 
Considering row	 2 column	 5 value	 0 
Considering row	 2 column	 1 value	 0 
Considering row	 2 column	 4 value	 0 
Considering row	 5 column	 1 value	 0 
Considering row	 5 column	 4 value	 0 
Considering row	 1 column	 4 value	 0.67 
integer(0)
> 
> # d.pizza example
> m <- cor(data.frame(lapply(d.pizza, as.numeric)), use="pairwise.complete.obs")
> FindCorr(m, verbose = TRUE)
Considering row	 8 column	 3 value	 0.018 
Considering row	 8 column	 2 value	 0.028 
Considering row	 8 column	 1 value	 0.03 
Considering row	 8 column	 12 value	 0.019 
Considering row	 8 column	 16 value	 0.076 
Considering row	 8 column	 5 value	 0.152 
Considering row	 8 column	 11 value	 0.095 
Considering row	 8 column	 13 value	 0.51 
Considering row	 8 column	 14 value	 0.478 
Considering row	 8 column	 6 value	 0.807 
Considering row	 8 column	 7 value	 0.543 
Considering row	 8 column	 9 value	 0.076 
Considering row	 8 column	 4 value	 0.042 
Considering row	 8 column	 10 value	 0.038 
Considering row	 8 column	 15 value	 0.033 
Considering row	 3 column	 2 value	 0.976 
  Flagging column	 3 
Considering row	 2 column	 1 value	 0.999 
  Flagging column	 2 
Considering row	 1 column	 12 value	 0.067 
Considering row	 1 column	 16 value	 0.072 
Considering row	 1 column	 5 value	 0.119 
Considering row	 1 column	 11 value	 0.056 
Considering row	 1 column	 13 value	 0.031 
Considering row	 1 column	 14 value	 0.017 
Considering row	 1 column	 6 value	 0.009 
Considering row	 1 column	 7 value	 0.01 
Considering row	 1 column	 9 value	 0.14 
Considering row	 1 column	 4 value	 0.038 
Considering row	 1 column	 10 value	 0.063 
Considering row	 1 column	 15 value	 0.015 
Considering row	 12 column	 16 value	 0.707 
Considering row	 12 column	 5 value	 0.292 
Considering row	 12 column	 11 value	 0.575 
Considering row	 12 column	 13 value	 0.05 
Considering row	 12 column	 14 value	 0.067 
Considering row	 12 column	 6 value	 0.043 
Considering row	 12 column	 7 value	 0.109 
Considering row	 12 column	 9 value	 0.072 
Considering row	 12 column	 4 value	 0.105 
Considering row	 12 column	 10 value	 0.003 
Considering row	 12 column	 15 value	 0.035 
Considering row	 16 column	 5 value	 0.227 
Considering row	 16 column	 11 value	 0.355 
Considering row	 16 column	 13 value	 0.077 
Considering row	 16 column	 14 value	 0.114 
Considering row	 16 column	 6 value	 0.008 
Considering row	 16 column	 7 value	 0.059 
Considering row	 16 column	 9 value	 0.248 
Considering row	 16 column	 4 value	 0.102 
Considering row	 16 column	 10 value	 0.045 
Considering row	 16 column	 15 value	 0.007 
Considering row	 5 column	 11 value	 0.478 
Considering row	 5 column	 13 value	 0.14 
Considering row	 5 column	 14 value	 0.12 
Considering row	 5 column	 6 value	 0.052 
Considering row	 5 column	 7 value	 0.013 
Considering row	 5 column	 9 value	 0.085 
Considering row	 5 column	 4 value	 0.111 
Considering row	 5 column	 10 value	 0.047 
Considering row	 5 column	 15 value	 0.01 
Considering row	 11 column	 13 value	 0.076 
Considering row	 11 column	 14 value	 0.082 
Considering row	 11 column	 6 value	 0.037 
Considering row	 11 column	 7 value	 0.014 
Considering row	 11 column	 9 value	 0.08 
Considering row	 11 column	 4 value	 0.046 
Considering row	 11 column	 10 value	 0.015 
Considering row	 11 column	 15 value	 0.011 
Considering row	 13 column	 14 value	 0.923 
  Flagging column	 13 
Considering row	 14 column	 6 value	 0.013 
Considering row	 14 column	 7 value	 0.009 
Considering row	 14 column	 9 value	 0.042 
Considering row	 14 column	 4 value	 0.016 
Considering row	 14 column	 10 value	 0.022 
Considering row	 14 column	 15 value	 0.021 
Considering row	 6 column	 7 value	 0.744 
Considering row	 6 column	 9 value	 0.037 
Considering row	 6 column	 4 value	 0.023 
Considering row	 6 column	 10 value	 0.006 
Considering row	 6 column	 15 value	 0.041 
Considering row	 7 column	 9 value	 0.034 
Considering row	 7 column	 4 value	 0.139 
Considering row	 7 column	 10 value	 0.032 
Considering row	 7 column	 15 value	 0.006 
Considering row	 9 column	 4 value	 0.252 
Considering row	 9 column	 10 value	 0.168 
Considering row	 9 column	 15 value	 0.005 
Considering row	 4 column	 10 value	 0.127 
Considering row	 4 column	 15 value	 0.011 
Considering row	 10 column	 15 value	 0.012 
[1]  3  2 13
> m[, FindCorr(m)]
                       week         date wine_ordered
index           0.974192573  0.999028828  0.030600322
date            0.976198358  1.000000000  0.036036580
week            1.000000000  0.976198358  0.032014141
weekday        -0.258535360 -0.042875700  0.013005712
area            0.091975225  0.120181706  0.140393613
count           0.010740354  0.005943085 -0.022125402
rabate          0.020039206 -0.010547837  0.013302530
price           0.018137168  0.028162980  0.509676944
operator        0.070854797  0.129699307  0.038239080
driver         -0.037057460 -0.066844465 -0.003367271
delivery_min    0.054344284  0.066614544  0.076473132
temperature     0.088823380  0.068222467 -0.049858606
wine_ordered    0.032014141  0.036036580  1.000000000
wine_delivered  0.016697141  0.020756359  0.922727399
wrongpizza      0.005659736  0.008217897  0.001967424
quality         0.099462346  0.080014155 -0.076622011
> 
> 
> 
> 
> 
> dev.off()
null device 
          1 
>


Created & Maintained by Osamu Ogasawara (osamu.ogasawara@gmail.com) and IMSBIO