Entering edit mode
3.9 years ago
QVINTVS_FABIVS_MAXIMVS
★
2.6k
Hi All I am a little lost at how to replicate this figure. I imagine the authors used corrplot in R to make it
With corrplot I am able to make a figure that has the size of the squares and color represent the correlation. However I am lost at making the figure where the size of the squares represent the p-value and the color represent the correlation.
I think they have either misunderstood or mislabeled the figure. The size always represents the correlation coefficient - 0 is smallest and 1 is largest
They don't seem to have a lot of values between 0 and 1, just 0 and -1. Their p-value legend is just plain wrong, IMO.
It's either that or they're correlating log(p-values), which would make no sense.
Yes it's confusing me quite a bit too. Thanks for the confidence in my sanity. They are not correlating pvalues but showing the pvalue of the strength of the pairwise correlation, I don't see how they could have made this plot because corrplot it seems you cannot separate the size of the square from the color (i.e. use two separate data sets: pvalues and correlations).
You are correct - size gives absolute magnitude of correlation and color gives signed magnitude. That p-value legend seems made up; it's not from the corrplot package. Corrplot doesn't do legends, just the blue-red sidebar.
yeah per the legend I imagine I could fabricate my own set of Pvalues to make it. I think the easiest thing to do perhaps is to hack the corrplot function to switch the variables for coloring