Victor D. Saldaña C.

PhD(c) in Geoinformatics Engineering
Technical University of Madrid (Spain)
June 2016

Overview.

This script allows to add to a group of ggplot2 plots laid out in panels with facets_grid

the values of the slope, intercept, R^2 and adjusted R^2 of every plot.

Context.

Let’s suppose you have a data frame with some interesting data and you want

to do some regression to fit a linear model with a straight line. Likewise, you want

to do the regression analysis by sub setting the data. In this case, ggplot2 package

offers a good set of tools to do it by dividing the date using “facets” that divides

the data frame according to the values of certain column. If the column you use to subset

the data frame have four values you will get four plots with its own regression line. Now,

if you want to add the regression equation and/or others values to compare them is

not possible with gggplot2, at least, in a direct way. Here is an option.

1. Load packages.

#1. Let's load the packages we need. If you haven't download them do it first.
library("ggplot2")
library("datasets")
library("plyr")

2. Create a function.

#2. We need to create a function. Let's call it "regression". It has only one argument, 
#the data frame (df). Likewise, let's use the "Toothgrowth" dataset from the package 
#“datasets”. This dataset contains 60 observations of "the Effect of Vitamin C on 
#Tooth Growth in Guinea Pigs". There are three variables: 
#1) Len (Tooth Length, numeric).
#2) Supp (Supplement Type, OJ = orange juice or VC = vitamin C, factor).
#3) Dose (Dose levels of vitamin, 0.5, 1, and 2 mg/day, numeric).

#2.1. Load the data and set a data frame (df).
data("ToothGrowth") #Load the dataset "ToothGrowth"
df<-ToothGrowth # Assigning it to a variables. 

#2.2. Create function.
regression=function(df){
        #setting the regression function. 
        reg_fun<-lm(formula=df$len~df$dose) #regression function
        #getting the slope, intercept, R square and adjusted R squared of 
        #the regression function (with 3 decimals).
        slope<-round(coef(reg_fun)[2],3)  
        intercept<-round(coef(reg_fun)[1],3) 
        R2<-round(as.numeric(summary(reg_fun)[8]),3)
        R2.Adj<-round(as.numeric(summary(reg_fun)[9]),3)
        c(slope,intercept,R2,R2.Adj)
        }

3.Subset the data frame.

#3. Now let's subset the data frame base on the values of the supplement type (supp) and
#let's apply the function "regression" to every data frame. To achieve this
#let's use the "ddply" function from the "plyr". The results in this case is another
#data frame (regressions_data) whose rows are the valeus of the slope, intercept, 
#R_squared and R_squared_adj for every data frame.
regressions_data<-ddply(ToothGrowth,"supp",regression)
colnames(regressions_data)<-c ("supp","slope","intercept","R2","R2.Adj")

4. Plotting.12

#4. Let's plot all the plots and the values of the regression models. 
qplot(dose, len, data = ToothGrowth, size=I(2))+geom_smooth(method="lm")+
        facet_grid(supp ~ .)+ggtitle("Regressions")+
        geom_label(data=regressions_data, inherit.aes=FALSE, aes(x = 1.2, y = 32,
label=paste("slope=",slope,","," ","intercept=",intercept,","," ","R^2=",R2,","," ","R^2.Adj=",R2.Adj)))

    1. If you don’t want a rectangle underneath the text use “geom_text” insted of “geom_label”.
    ↩
    1. is possible much more customization. Check out the help file of facets_grid, geom_label, etc.
    ↩