ggplot2_Densityplots.Rmd

---
output: html_document
---

```{r global_options, include=FALSE}
knitr::opts_chunk$set(warning=FALSE, message=FALSE, 
                      fig.width=5, fig.height=3, fig.align='center')
```

### ggplot2 Series 2 - Histogram and Density plot

```{r, warning = FALSE}
# Load libraries
library(ggplot2)
library(ggfortify)
library(dplyr)
library(knitr)
library(kableExtra)
```
The dataset we will use for this tutorial is from Kaggle: https://www.kaggle.com/c/house-prices-advanced-regression-techniques/data


```{r}
# Load data for demonstration
house_data <- read.csv('house_price.csv', header = TRUE)
options(scipen=999)  # turn off scientific notation like 1e+06
kable(head(house_data[c('GrLivArea', 'GarageType', 'SalePrice', 'CentralAir')], 3), align=c(rep('c',times=4))) %>%
      kable_styling(bootstrap_options=c("condensed"), full_width=FALSE)
```

Columns we will use in this tutorial:

 - GrLivArea: Above grade (ground) living area square feet
 - GarageType: Type of Garage location
 - SalePrice: the property's sale price in dollars
 - CentralAir - Central air conditioning (Binary): Y: yes, N: no

#### Basic Syntax
```{r}
#plot distribution of SalePrice
gg_hist <- ggplot(house_data, aes(x=SalePrice)) + geom_histogram()
print(gg_hist)
```


#### Customize histogram
```{r}
ggplot(house_data, aes(x=SalePrice)) + 
       geom_histogram(
                      alpha= .3,                  # set transparency of histogram
                      binwidth=50000,            # set the width of the bins
                      colour="lightsteelblue3",  # color of the outline
                      fill="lightsteelblue"      # color of bins
                     )
```


####  Add lines to the histogram
```{r}
ggplot(house_data, aes(x=SalePrice)) + 
       geom_histogram(alpha=.3, 
                      colour="lightsteelblue3", 
                      fill='lightsteelblue') + 
       geom_vline(xintercept=300000,             # vertical line
                  linetype="longdash",
                  color="lightsteelblue4", size=.5) +
       geom_hline(yintercept=100,                # horizontal line
                  linetype="dotdash",
                  color="lightsteelblue4", size=.5)
```

Some linetypes availables:
```{r pressure, echo=FALSE, fig.cap="Line types", out.width = '60%'}
include_graphics("line_type.JPG")
```

#### Add annotation to the histogram

```{r}
# Add mean and variance to the histogram
avg <- round(mean(house_data$SalePrice), 2)
std <- round(sd(house_data$SalePrice), 2)

# Method 1
ggplot(house_data, aes(x=SalePrice)) + 
       geom_histogram(alpha=.3, 
                       colour="lightsteelblue3", 
                       fill='lightsteelblue') + 
       annotate("text", x=350000, y=100, 
                     color="lightsteelblue4",
                     label=paste("Mean:", avg)) +
       annotate("text", x=350000, y=150, 
                 color="lightsteelblue4",
                 label=paste("Standard Deviaion:", std))

# Method 2: adding all annotations at once
annotation <- data.frame(
                         x=rep(350000, 2),
                         y=c(100, 150),
                         label=c(paste("Standard Deviaion:", std),
                                  paste("Mean:", avg))
                        )
ggplot(house_data, aes(x=SalePrice)) + 
       geom_histogram(alpha=.3, 
                       colour="lightsteelblue3", 
                       fill='lightsteelblue') + 
       geom_text(data=annotation, 
                      aes(x=x, y=y, label=label),                 
                      color="lightsteelblue4", 
                      size=4, angle=0, fontface="italic" )
```


#### Add kernel density line on to histogram

```{r}
# Add kernel density line on to histogram
ggplot(house_data, aes(x=SalePrice)) + 
       geom_histogram(aes(y=..density..),        # the y axis of histogram has to be density too
                      alpha=.3, 
                      colour="lightsteelblue3", 
                      fill='lightsteelblue') +
       geom_density(alpha=.3, 
                    color=NA,                    # no outline 
                    fill='lightsteelblue4')
```


#### Overlaid histograms
```{r}
# Two histograms seperated by categorical column: GarageType
ggplot(house_data, aes(x=SalePrice, 
                       fill=GarageType)) +       
       geom_histogram(alpha=.3, 
                      colour="lightsteelblue3") +
       scale_fill_brewer(palette="RdBu")         # change colour scheme

```

What if I want to overlay a few density plots? For example, I want to plot the density of SalePrice and GrLivArea at the same time. We can use the melt function from reshape2 package

```{r}
library(reshape2)
tmp <- house_data[c('TotalBsmtSF', 'GrLivArea')]
```

The melt function stacks a set of columns into a single column of data.
It generates a new dataframe is generated with two columns: variable and value

```{r}
kable(head(melt(tmp), 3), row.names = FALSE) %>%
      kable_styling(bootstrap_options=c("striped", "hover", "condensed"), full_width=FALSE)
kable(tail(melt(tmp), 3), row.names = FALSE) %>%
      kable_styling(bootstrap_options=c("striped", "hover", "condensed"), full_width=FALSE)

ggplot(melt(tmp), aes(x=value, fill=variable)) + 
       geom_histogram(aes(y=..density..), 
                      alpha=.3,
                      colour="lightsteelblue1") +
       geom_density(alpha=.5, 
                    color="lightsteelblue3") +
       scale_fill_brewer(palette="Pastel1") 
```

#### Add vertical lines to each of the histograms in the same plot
```{r}
# 1. Find the mean of each group
library(plyr)
mean_Ggroup <- ddply(house_data, "CentralAir", summarise, meanPrice=mean(SalePrice))
kable(mean_Ggroup, align=c(rep('c',times=2))) %>%
      kable_styling(bootstrap_options=c("striped", "hover", "condensed"), full_width=FALSE)

# 2. Add vertical lines using the mean value as xintercept
ggplot(house_data, aes(x=SalePrice, 
                       fill=CentralAir)) + # seperated by categorical column: GarageType
       geom_histogram(alpha=.3, 
                      colour="lightsteelblue3") +
       geom_vline(data=mean_Ggroup, 
                  aes(xintercept=meanPrice, colour=CentralAir),
                  linetype="dotdash", size=.5) +
       scale_fill_brewer(palette="Pastel1") # change colour scheme
```