Practical Predictive Analytics

Plotting the distributions

Creating a matrix plot of the variables is also a good preliminary step to perform, since it will immediately let you see the shape and distributions of the variables, and well as point out any gaps in the data. After removing some of the columns from the dataframe that we know we are not going to use, we can see that out dataset is fairly clean, and that the variables line up pretty much as we expect. For example, Admit.Day.of.Week is fairly normally distributed, but we can see that there is a null in the number of people discharged at mid-week. Costs are skewed, with very high values (but with low occurrences) at the extremes:

df <- df[,-c(12,13,16,19,20,21)]