plotting.Rmd

---
title: "Lesson 4 - basic R plotting"
author: "Mik Black"
date: "6 November 2015"
output: html_document
---

## R graphics

 - R provides multiple "systems" for producing graphics
 - We're going to focus on the oldest (and arguably easiest) of these: R base graphics
 - These capabilities are included with base R (i.e., what you get when you install the R software), and don't require any additional packages to be loaded.

---

## But first... data summaries

 - One of the reasons for generating plots is to produce a visual summary of the data
    - this helps us (and ohers) to understand key features such as _centre_ and _spread_.
 - In addition to graphical methods, we can also use tools to produce numerical summaries of this information.
    - For example: mean, median, standard deviation, variance, range, min, max, 
    interquartile range...

---

## Examples: non-graphical data summaries

```{r}
## Lets load the data and take a look at it first
data <- read.csv('dummy_data.csv')

## What type of object have we created?
class(data)

## What are its dimensions?
dim(data)

## What are the variable names?
names(data)

## What do the first few rows look like?
head(data)

## What does the start of the BMI variable look like?
## (note the use of the $ operator to access the BMI information from the data object)
head(data$BMI)

## What if we "attach" the data?
attach(data)

## Look, no more $ sign!
head(BMI)

## Summarize the entire data frame
summary(data)

## Summarize just the BMI variable
summary(BMI)

## Calculate the mean of the BMI variable
mean(BMI)

## Oops, we need to get rid of the NA values
mean(BMI,na.rm=T)

## Calculate the mean of the BMI variable
median(BMI, na.rm=T)

## Calculate the standard deviation of the BMI variable
sd(BMI, na.rm=T)
```

---

## Other summaries

 - the above are just a few summaries that can be produced.
    - try: IQR, max, min, range...
 - We can also summarize categorical data using tables via the ```table``` command.
    - ```prop.table``` can then be used to calculate proportions.

```{r}
## Make a table of the ethnicity data
## Note that information class labels can be a good thing...
table(ETHCLASS)

## Include the NA information
table(ETHCLASS, useNA = 'always')

## Diabetes status
table(DIABETES)

## Two-way table of ethnicity and diabetes
table(ETHCLASS, DIABETES)

## Add dimension names
table(ETHCLASS, DIABETES, dnn = c("Ethnicity","Diabetes"))

## We can also flip this via the transpose operator:
t( table(ETHCLASS, DIABETES) )

## Use the prop.table command to calculate proportions (cells sum to 1):
prop.table( table(ETHCLASS, DIABETES) )

## You can also calculate proportions across rows (rows sum to 1):
prop.table( table(ETHCLASS, DIABETES), 1 )

## or down columns (columns sum to 1):
prop.table( table(ETHCLASS, DIABETES), 2 )
```

---

## Plotting

 - Need to  use the right sort of plot for the data we have
    - One continuous variable: histogram (```hist```), boxplot (```boxplot```),
    density plot (```plot``` and ```density``` - see below)
    - Two continuous variabes: scatterplot (```plot```)
    - One (or more) cateogircal variables: barplots (```barplot```)
    - One continuous and one (or more) categorical variable: boxplots (```boxplot```)

---

## Histograms - single continuous variable

```{r}
## Generate histogram of age data
hist( AGECOL )
```

## Making plots prettier

 - most base R plotting methods share parameters relating to plot aesthetics (e.g., 
 axis labels/limits, title, text size, plotting colours etc)
 - this makes it very easy to customize plots, and may them look fairly nice

```{r}
## Add title and axis labels
hist( AGECOL , main = "Histogram of Age", xlab = "Age (years)")

##Add some colour...
hist( AGECOL , main = "Histogram of Age", xlab = "Age (years)", col='lightblue')

## Extend y axis
hist( AGECOL , main = "Histogram of Age", xlab = "Age (years)", ylim = c(0,400))
```

---

## Boxplots - single continuous variable

```{r}
## Boxplot of BMI
boxplot(BMI)
```

## Boxplots - continuous & categorical variables

```{r}
## Boxplot of BMI vs Diabetes Status
boxplot(BMI ~ DIABETES, xlab = "Diabetes status", ylab = "BMI", main = "BMI vs diabetes status")

## Change the colours and box widths
## Note: ```col``` defines the box colour, and ```border``` defines the border colour
boxplot(BMI ~ DIABETES, xlab = "Diabetes status", ylab = "BMI", main = "BMI vs diabetes status",
        boxwex = 0.2, col = "lightblue", border = "blue")
```

---

## Scatterplots - two continuous variables

```{r}
## Scatterplot of Age vs BMI
plot(AGECOL, BMI)

## Change plotting symbol, size and colour
plot(AGECOL, BMI, pch = 16, col = "blue", cex = 0.5)

## Use SEX variable to denote colour (has to be numeric, or actual colours)
## R colours are: 1 = black, 2 = red, 3 = green, 4 = blue, 5 = cyan, 6 = magenta,
##                7 = yellow, 8 = grey, 9 = black, 0 = white            
plot(AGECOL, BMI, col = SEX, pch = 16)

## Adjust y axis limits and add a legend
plot(AGECOL, BMI, col = SEX, pch = 16, ylim = c(18, 80))
legend(20, 80, c("Male", "Female"), fill = c("black","red"))
```

---

## Barplots - one or more categorical variables

```{r}
## Summarize ethinicty data as a table:
table(ETHCLASS)

## Represent this information as a barplot
barplot( table(ETHCLASS) )

## Generate 2-way tabe for ethnicity and diabetes
table(ETHCLASS, DIABETES)

## Represent this information as a (stacked) barplot
barplot( table(ETHCLASS, DIABETES), xlab = "Diabetes status", ylab = "Frequency" )

## Same information, unstacked
barplot( table(ETHCLASS, DIABETES), xlab = "Diabetes status", ylab = "Frequency", beside = TRUE )

## What is we want to look at it the other way around?
## Use the transpose function, t(), to flip the table
t( table(ETHCLASS, DIABETES) )

## Then make a barplot
barplot( t(table(ETHCLASS, DIABETES)), beside = TRUE )

## What if we're more interested in proportions?
prop.table( table(ETHCLASS, DIABETES) )

## Make the corresponding barplot
barplot( prop.table(table(ETHCLASS, DIABETES)) )

## Or we can make the columns sum to one
prop.table( table(ETHCLASS, DIABETES), 2)
barplot( prop.table(table(ETHCLASS, DIABETES), 2) )
```

---

## Density plots

 - Sometines it is useful to represent continuous data as a density, rather than
 a histogram.
 
```{r}
## Histogram of BMI data 
hist(BMI)

## Density plot of BMI data
## Note: have to remove NA values, or the density() function gives an error.
plot( density(BMI, na.rm = T) )
```

---

## Multiple plots per page

 - Arranging multiple plots on a single figure can be useful
 - In base R graphics this is accomplished using either the ```mfrow()``` or
 ```mfcol()``` functions.
   - both specify a grid of plots: which function you use depends on the order in which you want to generate the plots.
   - ```mfrow()``` fills the grid row-by-row (left to right)
   - ```mfcol()``` fills the grid column-by-column (top to bottom)
   
```{r}
## Specify a 1 x 2 (1 row, 2 columns) grid of plots, and fill by rows, left to right
par(mfrow = c(1, 2))

## Place a histogram of BMI in the first cell of the first row
hist(BMI)

## Place a density plot of BMI in the second cell of the first row
plot(density(BMI, na.rm=T))
```

---

## Saving plots

 - R can output graphics in a number of formats: pdf, jpeg, png etc
 - In RStudio you can also save your plots via the "Export" menu in the "Plots" tab.
 - The following code opens a PDF file, writes a histogram to that file, and then
 closes the connection to the file.
 
```{r}
## Open PDF file for writing, and specify dimensions
## Note that PDF files use dimensions based on inches, whereas png and 
## jpeg files use pixels
pdf(file = 'bmi_histogram.pdf', height = 6, width = 6)

## Write histogram of BMI to file
hist(BMI)

## Close the file
dev.off()

## For multi-plot figures, you need to adjust the dimensions
pdf(file = 'bmi_histogram_density.pdf', height = 6, width = 12)
par(mfrow = c(1, 2))
hist(BMI)
plot(density(BMI, na.rm=T))
dev.off()
```