MULTIVARIATE DATA
Looking at multivariate data is not that much different than examining bivariate data. This
section explores how we expand our notion of some R functions for multivariate data.
n-way Contingency Tables
These are for categorical data and are no different from the contingency tables in Chapter 3,
except that there are more dimensions (variables) to the table. The text covers this well on its
own. The function
ftable( )
is particularly useful.
> library(MASS); data(Cars93); attach(Cars93)
> data.frame(Make, MPG.highway, Price)
Using the
cut()
function, we can divide numeric data into intervals.
> price = cut(Price,c(0,12,20,max(Price)))
> levels(price)=c("cheap","okay","expensive")
> mpg = cut(MPG.highway,c(0,20,30,max(MPG.highway)))
> levels(mpg) = c("poor","decent","excellent")
Let’s look at some tables. Are there any apparent patterns in the 3-way contingency table?
> table(Type)
> table(price,Type)
> table(price,Type,mpg)
> ftable(price,Type,mpg)
> ftable(price,Type,mpg, col.vars=c(‘price’,’mpg’))
We can even use our new variable
price
to make barplots.
> barplot(table(price,Type),beside=T) # the price by different types
> barplot(table(Type,price),beside=T) # type by different prices
Independent Samples
As discussed earlier, sometimes we are interested in obtaining the same variable data from
several independent sample sources. The text gives an excellent example when the samples each
have their own variable name. But what about a case when the samples are their own variable.
Look at the
PlantGrowth
dataset.
> data(PlantGrowth)
> PlantGrowth[1:5, ]
There are 3 groups a control and two treatments. For each group, weights are recorded. The data
is generated this way, by recording a weight and group for each plant. However, you may want
to plot boxplots for the data broken down by their group. How to do this?
