How to correlate all columns on dataframe and build a heatmap?

Question

I have a dataframe with my measurements separated by days of the month, and I would like to build a correlation matrix (heatmap) between the months like the example below:

Thank you for the help! The head of my data is:

> head(month)
          Jan         Feb          Mar          Apr          May
1 18.06474972   11.399055 -31.03085327  127.2022133 -3460.850586
2 51.03550959 461.5850525  95.19575119  243.0290451  -2083.35199
3 47.54027748 110.2497387 -74.04249191  546.8690033 -1276.771118
4 340.8937378 282.1457062  74.03128743  322.6983643 -203.3483124
5 247.5398369 30.00830674  370.5808411  416.0554199  121.7065811
6 399.5107269 922.6151428  439.3453064 -35.20425129 -1469.418518
           Jun         Jul         Aug          Sep          Oct
1  39.49685574 2622.950073 1649.752869   1238.56366  500.0726166
2  380.0546722 2552.876099 2572.219482  525.0535278  1092.828491
3  278.1513977 1118.665558  2370.05896  561.7017517  880.8520813
4  1334.320251 1658.281433 588.7766266  316.3756104   331.518219
5  681.0978546 1830.663025 436.6187897  188.4522657 -106.0038891
6 -69.24006653 1645.978577 2079.013062 -116.1453476  292.2966232
           Nov          Dec
1 -251.1712646  196.3089867
2 -2490.553711  37.59177971
3 -1460.746155  251.9935303
4  425.8812321  345.0469666
5  253.5491486 -150.1276989
6  2654.713806 -1673.095764

StupidWolf · Accepted Answer · 2021-03-18 14:31:26Z

4

You can try something like below:

library(pheatmap)
month = data.frame(matrix(runif(120),ncol=12))
colnames(month) = format(ISOdate(2004,1:12,1),"%B")

pheatmap(cor(month),cluster_rows = FALSE, cluster_cols =FALSE)

answered Mar 18, 2021 at 14:31

StupidWolf

47.1k17 gold badges50 silver badges81 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Krzysztof Over a year ago

While plotting correlations please don't forget about the color scale. It can get confusing to interpret the results.

library(RColorBrewer); pheatmap(cor(month),cluster_rows = FALSE, cluster_cols=FALSE, color=colorRampPalette(rev(brewer.pal(n = 7, name = "RdYlBu")))(length(seq(-1, 1, 0.01))), breaks=seq(-1, 1, 0.01))

G. Grothendieck · Accepted Answer · 2021-03-18 19:40:36Z

The corrplot package has numerous formats for displaying correlation plots. Here are a few. example(corrplot) and example(corrplot.mixed) show other possibilities.

As in the question, the second plot:

uses the color maroon for +1 and a light blue for -1 using the pals package for its palettes (RColorBrewer is another possibility), and
places the numeric value of the correlation on each tile

but you can change the color scheme and/or format as desired.

library(corrplot)
library(pals)

opar = par(mfcol = 1:2)
wat <- adjustcolor(watlington(16)[11:3], alpha = 0.7)
corrplot.mixed(cor(month), lower = "number", upper = "shade", upper.col = wat)
corrplot(cor(month), col = wat, cl.length = 21, 
  addCoef.col = "black", method = "shade")
par(opar)

Note

We have used the first 5 months since the data was not provided in an easily usable form in the question but it should work the same with all 12 months.

Lines <- "          Jan         Feb          Mar          Apr          May
1 18.06474972   11.399055 -31.03085327  127.2022133 -3460.850586
2 51.03550959 461.5850525  95.19575119  243.0290451  -2083.35199
3 47.54027748 110.2497387 -74.04249191  546.8690033 -1276.771118
4 340.8937378 282.1457062  74.03128743  322.6983643 -203.3483124
5 247.5398369 30.00830674  370.5808411  416.0554199  121.7065811
6 399.5107269 922.6151428  439.3453064 -35.20425129 -1469.418518"
month <- read.table(text = Lines)

Collectives™ on Stack Overflow

How to correlate all columns on dataframe and build a heatmap?

2 Answers 2

1 Comment

Note

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Note

Comments

Your Answer

Sign up or log in

Post as a guest

Related