Indicating the statistically significant difference in bar graph base R

Question

This has been asked before in this post: Indicating the statistically significant difference in bar graph USING R. However, they wanted to know how to do this using ggplot2. I was wondering how you do this using just the base package or function barplot(). I want something that looks like this image below:

https://i.stack.imgur.com/3I6El.jpg

my current code:

barcenter3<- barplot(newMEANs3$Percent_Viability, names.arg=c("Control", "Cyp28d1", "A3", "A4"), ylab = "Average Emergent",  ylim=c(0, 1.1), xlab= "RNAi Line", main = "Trip Nicotine UAS-RNAi Emergents")
segments(barcenter3, newMEANs3$Percent_Viability-newSDs3$Percent_Viability, barcenter3, newMEANs3$Percent_Viability+newSDs3$Percent_Viability, lwd=1);
segments(barcenter3 - 0.1, newMEANs3$Percent_Viability-newSDs3$Percent_Viability, barcenter3  + 0.1, newMEANs3$Percent_Viability-newSDs3$Percent_Viability, lwd=1);
segments(barcenter3 - 0.1, newMEANs3$Percent_Viability+newSDs3$Percent_Viability, barcenter3  + 0.1, newMEANs3$Percent_Viability+newSDs3$Percent_Viability, lwd=1);
dev.off();

I want to add p value comparing contrast.

So you want to know how to do that image in base graphics? Did you try anything yet? — Mike Wise, Jul 02 '15 at 18:40
I can only get to the point of making barplot with error bars. — user3919708, Jul 02 '15 at 18:42
Mike I added my current code that is all I have. I dont know where to go from here. — user3919708, Jul 02 '15 at 18:44
Excellent. Thank you. Will have a look at it as soon as I finish this other task (if it is still open) :) — Mike Wise, Jul 02 '15 at 18:45
Still no data so the code it not run-able. See [how to make a reproducible example](http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example). — MrFlick, Jul 02 '15 at 18:56

Rorschach · Accepted Answer · 2015-07-02T19:32:10.013

Here is a simple function to do this.

## Sample Data
means <- seq(10,40,10)
pvals <- seq(0.01, 0.05, 0.02)

barPs <- function(means, pvals, offset=1, ...) {
    breaks <- barplot(means, ylim=c(0, max(means)+3*offset), ...)
    ylims <- diff(means) + means[-length(means)] + offset
    segments(x0=breaks[-length(breaks)], y0=ylims, x1=breaks[-1], y1=ylims)
    segments(x0=c(breaks[-length(breaks)], breaks[-1]),
             y0=rep(ylims, each=2), y1=rep(ylims-offset/2, each=2))
    text(breaks[-length(breaks)]+diff(breaks[1:2])/2, ylims+offset, 
         labels=paste("p=", pvals))
}

barPs(means, pvals, offset=1, main="Bar w/ P-value", 
      names.arg=toupper(letters[1:4]))

enter image description here

The segments that make the small lines for the brackets only work for the case of 4 bars like this! If you want it to work with more bars as well, the numbers need to be sorted! `segments(x0= sort(c(breaks[-length(breaks)], breaks[-1])), ...` — bug313, Mar 01 '22 at 15:11

Indicating the statistically significant difference in bar graph base R

1 Answers1