So my advisor pointed out this ’new’ (well, 2004), way of plotting results of logistic regression results. The idea was presented in a 2004 Bulletin of the Ecological Society of America issue (here). I tried to come up with a solution using, what else, ggplot2. I don’t have it quite all the way down - I am missing the second y-axis values for the histograms, but someone smarter than me can figure that part out (note that Hadley doesn’t want to support second y-axes in ggplot2, but they can probably be hacked on).
Here’s the code (originally was in https://gist.github.com/1589136):
# Define the function
loghistplot <- function(data) {
require(ggplot2); require(gridExtra) # load packages
names(data) <- c('x','y') # rename columns
# get min and max axis values
min_x <- min(data$x)
max_x <- max(data$x)
min_y <- min(data$y)
max_y <- max(data$y)
# get bin numbers
bin_no <- max(hist(data$x)$counts) + 5
# create plots
a <- ggplot(data, aes(x = x, y = y)) +
theme_bw(base_size=16) +
geom_smooth(method = "glm", family = "binomial", se = TRUE,
colour='black', size=1.5, alpha = 0.3) +
# scale_y_continuous(limits=c(0,1), breaks=c(0,1)) +
scale_x_continuous(limits=c(min_x,max_x)) +
opts(panel.grid.major = theme_blank(),
panel.grid.minor=theme_blank(),
panel.background = theme_blank()) +
labs(y = "Probability\n", x = "\nYour X Variable")
b <- ggplot(data[data$y == unique(data$y)[1], ], aes(x = x)) +
theme_bw(base_size=16) +
geom_histogram(fill = "grey") +
scale_y_continuous(limits=c(0,bin_no)) +
scale_x_continuous(limits=c(min_x,max_x)) +
opts(panel.grid.major = theme_blank(),
panel.grid.minor=theme_blank(),
axis.text.y = theme_blank(),
axis.text.x = theme_blank(),
axis.ticks = theme_blank(),
panel.border = theme_blank(),
panel.background = theme_blank()) +
labs(y='\n', x='\n')
c <- ggplot(data[data$y == unique(data$y)[2], ], aes(x = x)) +
theme_bw(base_size=16) +
geom_histogram(fill = "grey") +
scale_y_continuous(trans='reverse') +
scale_y_continuous(trans='reverse', limits=c(bin_no,0)) +
scale_x_continuous(limits=c(min_x,max_x)) +
opts(panel.grid.major = theme_blank(),panel.grid.minor=theme_blank(),
axis.text.y = theme_blank(), axis.text.x = theme_blank(),
axis.ticks = theme_blank(),
panel.border = theme_blank(),
panel.background = theme_blank()) +
labs(y='\n', x='\n')
grid.newpage()
pushViewport(viewport(layout = grid.layout(1,1)))
vpa_ <- viewport(width = 1, height = 1, x = 0.5, y = 0.5)
vpb_ <- viewport(width = 1, height = 1, x = 0.5, y = 0.5)
vpc_ <- viewport(width = 1, height = 1, x = 0.5, y = 0.5)
print(b, vp = vpb_)
print(c, vp = vpc_)
print(a, vp = vpa_)
}
# Examples
# loghistplot(mtcars[,c("mpg","vs")])
# loghistplot(movies[,c("rating","Action")])
logpointplot <- function(data) {
require(ggplot2); require(gridExtra) # load packages
names(data) <- c('x','y') # rename columns
# get min and max axis values
min_x <- min(data$x)
max_x <- max(data$x)
min_y <- min(data$y)
max_y <- max(data$y)
# create plots
ggplot(data, aes(x = x, y = y)) +
theme_bw(base_size=16) +
geom_point(alpha = 0.5, position = position_jitter(w=0, h=0.02)) +
geom_smooth(method = "glm", family = "binomial", se = TRUE,
colour='black', size=1.5, alpha = 0.3) +
scale_x_continuous(limits=c(min_x,max_x)) +
opts(panel.grid.major = theme_blank(),
panel.grid.minor=theme_blank(),
panel.background = theme_blank()) +
labs(y = "Probability\n", x = "\nYour X Variable")
}
# Examples
# logpointplot(mtcars[,c("mpg","vs")])
# logpointplot(movies[,c("rating","Action")])
Here’s a few examples using datasets provided with the ggplot2 package:
loghistplot(mtcars[,c("mpg","vs")])
loghistplot(movies[,c("rating","Action")])
And two examples of the logpointplot function:
logpointplot(mtcars[,c("mpg","vs")])
logpointplot(movies[,c("rating","Action")])
