9

How can I add RMSE, slope, intercept and r^2 to a plot using R? I have attached a script with sample data, which is a similar format to my real dataset--unfortunately, I am at a stand-still. Is there an easier way to add these statistics to the graph than to create an object from an equation and insert that into text()? I would ideally like the statistics to be displayed stacked on the graph. How can I accomplish this?

## Generate Sample Data
x = c(2,4,6,8,9,4,5,7,8,9,10)
y = c(4,7,6,5,8,9,5,6,7,9,10)

# Create a dataframe to resemble existing data
mydata = data.frame(x,y)

#Plot the data
plot(mydata$x,mydata$y)
abline(fit <- lm(y~x))

# Calculate RMSE
model = sqrt(deviance(fit)/df.residual(fit))

# Add RMSE value to plot
text(3,9,model)
Borealis
  • 8,044
  • 17
  • 64
  • 112
  • Look at this question http://stackoverflow.com/questions/7549694/ggplot2-adding-regression-line-equation-and-r2-on-graph – mnel Oct 29 '12 at 00:37
  • Any ideas on how to add "RMSE = value" in addition to r^2 and an eq for slope and intercept in a neat, formatted fashion? – Borealis Oct 29 '12 at 00:41
  • You haven't provided any details about what you mean by "formatted fashion". In base graphics, the way you add text to a plot is with `text` (or `mtext`), and you got a link to an excellent answer for ggplot. What problem remains? – joran Oct 29 '12 at 00:47
  • You might also look at the examples in `?plotmath` for some ideas. – joran Oct 29 '12 at 00:58

1 Answers1

18

Here is a version using base graphics and ?plotmath to draw the plot and annotate it

## Generate Sample Data
x = c(2,4,6,8,9,4,5,7,8,9,10)
y = c(4,7,6,5,8,9,5,6,7,9,10)

## Create a dataframe to resemble existing data
mydata = data.frame(x,y)

## fit model
fit <- lm(y~x, data = mydata)

Next calculate the values you want to appear in the annotation. I prefer bquote() for this, where anything marked-up in .(foo) will be replaced by the value of the object foo. The Answer @mnel points you to in the comments uses substitute() to achieve the same thing but via different means. So I create objects in the workspace for each value you might wish to display in the annotation:

## Calculate RMSE and other values
rmse <- round(sqrt(mean(resid(fit)^2)), 2)
coefs <- coef(fit)
b0 <- round(coefs[1], 2)
b1 <- round(coefs[2],2)
r2 <- round(summary(fit)$r.squared, 2)

Now build up the equation using constructs described in ?plotmath:

eqn <- bquote(italic(y) == .(b0) + .(b1)*italic(x) * "," ~~ 
                  r^2 == .(r2) * "," ~~ RMSE == .(rmse))

Once that is done you can draw the plot and annotate it with your expression

## Plot the data
plot(y ~ x, data = mydata)
abline(fit)
text(2, 10, eqn, pos = 4)

Which gives:

enter image description here

Gavin Simpson
  • 170,508
  • 25
  • 396
  • 453