08-multiple_plots.Rmd

# Multiple plots 

```{r, echo=FALSE}
sapdecay <- read.table(
     "http://www.bio.ic.ac.uk/research/mjcraw/therbook/data/sapdecay.txt", 
     header=T
     )
```

You may have noticed in one of the previous examples, that two plots were produced one on top of the other. This is a pretty easy thing to accomplish.

All we need to do is split the plotting area with the <tt>par(mfrow=c(x,y))</tt> command. As with subscripts, with this command <tt>x</tt> codes for the number of rows, and <tt>y</tt> represents the number of columns. Check <tt>?par</tt> for more information.


```{r multiple_plots-1}

par(mfrow = c(2,2)) 

with(
     sapdecay,
     plot(
          x = x,
          y = y,
          xlab = "Time (hours)",
          ylab = "Decay rate",
          type = "l",
          col = "red",
          lwd = 2,
          main = "Line"
          )
     )

with(
     sapdecay,
     plot(
          x = x,
          y = y,
          xlab = "Time (hours)",
          ylab = "Decay rate",
          type = "o",
          col = "green",
          lwd = 1,
          main = "Overplot"
          )
     )

with(
     sapdecay,
     plot(
          x = x,
          y = y,
          xlab = "Time (hours)",
          ylab = "Decay rate",
          type = "h",
          col = "blue",
          lwd = 2,
          main = "Histogram"
          )
     )

with(
     sapdecay,
     plot(
          x = x,
          y = y,
          xlab = "Time (hours)",
          ylab = "Decay rate",
          type = "s",
          col = "hotpink",
          lwd = 2,
          main = "Step"
          )
     )

# It's important that we reset the mfrow parameter at the end of plotting, 
# otherwise all subsequent plots will appear on this four by four grid, which
# may not be what we want.

par(mfrow = c(1,1))

```

Using this method, it's also perfectly possible to combine plots of two different types, and plot them next to each other.

Let's imagine that we want to test whether a variable is normally distributed - there are two very simple graphical commands built into R that make this easier: <tt>qqnorm</tt> and <tt>qqline</tt>. These commands produce quantile-quantile plots for the normal distribution. Essentially if the data are normally distributed, they will conform to the straight line<a href="#footnote-1">[1]</a>.

First we will create a vector of random numbers drawn from a normal distribution - so we already know that they are normal!

```{r ,echo=-1,}
set.seed(1337)
normal_data <- rnorm(
     100, 
     mean=0, 
     sd=1
     )

normal_data
```

And now...some plots. First, two plots side by side.

```{r multiple_plots-2, fig.width=12}

par(
     mfrow=c(1,2)
     ) 

# Remember that the first number is the number of rows, the latter is the number
# of columns. So here, we have specified one row, two columns...

# Now the plots:

hist(normal_data)

qqnorm(normal_data)
qqline(normal_data)
```

So things should look pretty normal. In this example we have not specified a 'seed' (see <tt>?set.seed</tt> for more info). So every time that you run the code, you will get a different set of randomly generated numbers drawn from a normal distribution.

All well and good, but what about one above the other? Just changed the 1 and 2 around in the <tt>par(mfrow=c(x,y))</tt> command.

```{r multiple_plots-3, fig.height=12}

par(mfrow=c(2,1)) 

# Remember that the first number is the number of rows, the latter is the number
# of columns. So here, we have specified one row, two columns...

# Now the plots:

hist(normal_data)

qqnorm(normal_data)
qqline(normal_data)
```

<div class="ex">

### Exercise 

Using what you have learnt in the previous sections, and the data from the <tt>car_prices</tt>, have a go at reproducing the plots you see below. Most of what you need to know has been covered, though you will need to consult the help files for certain commands.
</div>

```{r multiple_plots-4, fig.width=15, fig.height=15}

car_prices <- read.table(
  "discount.csv", 
  header = T,
  sep = ","
)

par(
  mar = c( 4.1, 4.5, 0.5, 0.5),
  mgp = c( 2.1, 0.6, 0),
  mfrow = c(2,2),
  cex = 1.4,
  cex.lab = 1.2,
  cex.axis = 1.2,
  lend="square"
)

diesel <- subset(car_prices,fuel == "diesel")
petrol <- subset(car_prices,fuel == "petrol")

p1<-hist(
  diesel$price,
  ylim = c(0,60),
  col = rgb(1,0,0,0.5), # Note that the colours used must be those that allow transparency
  main = "",
  breaks = 8,
  xlab = bquote(Asking~price~(10^3~"£")),
  xlim = c(5,5500),
  yaxt = "n",
  xaxt = "n"
)

# plot second graph - note use of 'add = TRUE' to overlay onto the first plot

p2<-hist(
  petrol$price,
  breaks = 8,
  add = TRUE,
  col = rgb(0,0,1,0.5)
)

axis(
  side = 2,
  at = seq(0,60,10),
  las = 2
)

axis(
  side = 1,
  labels = seq(0,5,1),
  at = seq(0,5000,1000),
  las = 1
)

legend(
  "topleft",
  legend = c("Petrol","Diesel"),
  fill = c(rgb(0,0,1,0.5),rgb(1,0,0,0.5)),
  bty = "n"
)



boxplot(
  split(
    car_prices$price,
    car_prices$fuel
  ),
  col = c(rgb(1,0,0,0.5),rgb(0,0,1,0.5)),
  ylab = bquote(Asking~price~(10^3~"£")),
  yaxt = "n"
)

axis(
  side = 2,
  labels = seq(0,10,1),
  at = seq(0,10000,1000),
  las = 2
)



with(
  petrol,
  plot(
    miles,
    price,
    col = rgb(0,0,1,0.5),
    ylim = c(500,5000),
    xlim=c(20000,180000),
    xlab = bquote(Mileage~(10^3~miles)),
    ylab = bquote(Asking~price~(10^3~"£")),
    yaxt = "n",
    xaxt = "n",
    bty = "l",
    pch = 16,
    cex = 1.2
  )
)

with(
  diesel,
  points(
    miles,
    price,
    col = rgb(1,0,0,0.5),
    pch = 17,
    cex = 1.2
  )
)

axis(
  side = 2,
  labels = seq(0,10,1),
  at = seq(0,10000,1000),
  las = 2
)

axis(
  side = 1,
  at = seq(0,200000,20000),
  labels = seq(0,200,20)
)

legend(
  "topright",
  col = c(rgb(0,0,1,0.5),rgb(1,0,0,0.5)),
  legend = c("Petrol","Diesel"),
  pch = c(16:17),
  bty = "n",
  cex = 1.2
)



###############

qqnorm(
  car_prices$miles,
  main = "QQ plot of mileage"
)

qqline(
  car_prices$miles,
  col = "red",
  lty = 2
)

# boxplot(
#      split(
#           car_prices$miles/car_prices$price,
#           car_prices$fuel
#           ),
#      col = c(rgb(1,0,0,0.5),rgb(0,0,1,0.5)),
#      ylab = bquote(Miles~"£"^-1),
#      yaxt = "n"
#      )

# axis(
#      side = 2,
#      at = seq(0,200,40),
#      las = 2
# )

par(mfrow=c(1,1))

```


< id="footnote-1">[1] I'm not going to dwell on the reasons why this is so - we're more interested in the plotting graphs than the statistics behind it in this sections, but if you don't understand what is going on here, you should look it up! <a href="http://en.wikipedia.org/wiki/Q%E2%80%93Q_plot">http://en.wikipedia.org/wiki/Q-Q_plot</a>.