How to create a plot of a categorical variable versus a continuous (numerical one) and deal with overploting.

ggplot_diamonds_exploration.R

library(ggplot2)

# some background about the data
# http://www.diamondse.info/
?diamonds
summary(diamonds)
str(diamonds)
is.factor(diamonds$clarity)
is.factor(diamonds$color)
# clarity and color are already factor variables
# In many cases you will need to convert categorical variables to factors 
# when you read in your own dataset.

ggplot(aes(x=color,y=price), data=diamonds) + geom_point()

# let's try jittering
?geom_jitter

ggplot(aes(x=color,y=price), data=diamonds) + 
  geom_jitter(position=position_jitter(width=0.5, height=0))

ggplot(aes(x=color,y=price), data=diamonds) + 
  geom_jitter(position=position_jitter(width=0.3, height=0), alpha=0.1)

# we might need to subsample - ~3000 observations
sample_data <- function(sample_size=3000) {
  index <- sample(nrow(diamonds), sample_size)
  return(diamonds[index, ])
}

# use a sample of the data and add box-plots
ggplot(aes(x=color,y=price), data=sample_data()) + 
  geom_boxplot(outlier.size=0) + 
  geom_jitter(position=position_jitter(width=0.3, height=0), alpha=0.4, 
              color="steelblue2")

# we will add a third variable - clarity
ggplot(aes(x=color,y=price, color=clarity), data=sample_data()) + 
  geom_jitter(position=position_jitter(width=0.3, height=0), alpha=0.8)

# Instead of using color to distringuish different levels of 
# clarity, we can use facets.
# We need to increase our sample of the data in order to 
# have enough data for each facet in our plot

ggplot(aes(x=color,y=price), data=sample_data(5000)) +
  geom_boxplot(outlier.size=0, color="blue") +
  geom_jitter(position=position_jitter(width=0.3, height=0), alpha=0.2) +
  facet_wrap(~clarity)

# not a perfect plot - what are ways in which we can improve it?
# It shows trade-offs between showing box plots and the
# distributions of elements