venerdì 19 marzo 2010

Balloon plot using ggplot2

Following Tal Galili example and using part of his code, I want to plot the balloonplot you can see here using R and the excellent ggplot2 package by Hadley Wickham.

### I retrieve the data from the google document you can find here using Tal Galili code:
## I slightly modified Tal code to include popularity stats:
supplement.popularity <- supplements.data[ss,7]
supplements.df <- na.omit(data.frame(supplement.name, supplement.benefits, supplement.popularity, supplement.score)) ## remove rows containing NAs
colnames(supplements.df) <- c("name", "benefits", "popularity", "score")
## For sake of simplicity I select only the cardio metacondition
cardio <- (supplements.df[supplement.benefits=="cardio",])[, -2]
## For reproducibility I add the cardio data.frame so you can use it right away
cardio <- read.table(tc <-textConnection(
" name popularity score
2 'arginine' 1.080 3
10 'vitamin b3' 0.201 3
15 'omega 3' 4.000 3
22 'hawthorn' 0.442 4
27 'red yeast rice' 0.264 4
29 'vitamin d' 6.700 4
31 'omega 6' 2.000 4
35 'green tea' 26.100 5
37 'olive leaf' 0.224 5
41 'fish oil' 4.000 6
43 'red yeast rice' 0.264 6")); close(tc)
cardio$name <- gsub(" ", "\n", cardio$name) #substitute ' ' with '\n' in the names
library(ggplot2)
myTheme <- function(base_size = 10) {
structure(list(
panel.background = theme_rect(size = 1, colour = "lightgray"),
panel.grid.major = theme_blank(),
panel.grid.minor = theme_blank(),
axis.line = theme_blank(),
axis.text.x = theme_blank(),
axis.ticks = theme_blank(),
strip.background = theme_blank(),
strip.text.y = theme_blank(),
legend.background = theme_blank(),
legend.key = theme_blank(),
legend.key.size = unit(1.2, "lines"),
legend.title = theme_text(size = 8, face = "bold", hjust = 0),
legend.position = "right"
), class = "options")
}
s <- ggplot(cardio, aes(name, score)) + xlab(NULL) + ylab(NULL) + myTheme()
s <- s + geom_point( aes(size=popularity, colour=score, fill=score), legend=TRUE) +
scale_y_continuous( breaks=as.numeric(levels(factor(cardio$score))), labels=c("Conflicting", "Promising", "Good", "Strong") ) +
scale_area( breaks=c(min(cardio$popularity),mean(cardio$popularity),max(cardio$popularity)), to=c(4,60) ) +
geom_text(aes(y=cardio$score, label=cardio$name, size=cardio$popularity/90), legend=FALSE)
#pdf("cardio.pdf",height=8,width=12);s;dev.off()
png("cardio.png",height=700,width=1000);s;dev.off()

domenica 7 marzo 2010

One R Tip A Day meets Tecnica Arcana

For italian speaking people only (sorry!).

Carlo il curatore dell'ottimo podcast tecnologico Tecnica Arcana mi ha intervistato sulla mia professione e su R. Qui potete scaricare l'intervista in formato mp3.