Machine Learning Made Easy

Density Plots

Introduction

Density plots are a great way to understand the distribution of the variable. This can be used to get an idea of how different levels of a categorical variable impacts a numeric variable.

Installing the library: dplyr,tidyr and Ecdat package

package.name<-c("dplyr","tidyr","Ecdat","ggplot2")

for(i in package.name){

  if(!require(i,character.only = T)){

    install.packages(i)
  }
  library(i,character.only = T)

}


# Ecdat package has the 'Health Insurance and Hours Worked By Wives' data
data(HI)
df<-HI
head(df)

  whrswk hhi whi hhi2  education  race hispanic experience kidslt6 kids618
1      0  no  no   no 13-15years white       no       13.0       2       1
2     50  no yes   no 13-15years white       no       24.0       0       1
3     40 yes  no  yes    12years white       no       43.0       0       0
4     40  no yes  yes 13-15years white       no       17.0       0       1
5      0 yes  no  yes  9-11years white       no       44.5       0       0
6     40 yes yes  yes    12years white       no       32.0       0       0
   husby       region   wght
1 11.960 northcentral 214986
2  1.200 northcentral 210119
3 31.275 northcentral 219955
4  9.000 northcentral 210317
5  0.000 northcentral 219955
6 15.690 northcentral 208148

Step 1:Lets look at spread of ‘husby’(husband’s income) against ‘region’

An important step here is to convert the categorical variable into factor

interim.df<-df%>%
  select(region,husby)

interim.df$region<-as.factor(interim.df$region)

Step 2:Density Plot to look at distirbution of experience cut across race

p <- ggplot(interim.df, aes(x=husby)) + 
  geom_density(color="darkblue", fill="lightblue")+
  geom_vline(aes(xintercept=mean(husby)),
            color="blue", linetype="dashed", size=1)

plot(p)

We can play around with ‘fill’ of the plot when considering only a single plot.When looking at multiple plots, ‘fill’ attribute will make it difficult to interpret the plot

mean.df<-interim.df%>%
  group_by(region)%>%
  summarise(gp.mean=mean(husby))

p<-ggplot(interim.df, aes(x=husby, color=region)) +
  geom_density()+
  geom_vline(data=mean.df, aes(xintercept=gp.mean, color=region),
             linetype="dashed")
plot(p)

Final Comments

We can see that the regions dont vary that much in terms of husband’s income. Hence it can be said that regions are similar as far as this indiciator is concerned