Data Science and Statistical Computing

class: center, middle, inverse, title-slide

.title[
# Data Science and Statistical Computing
]
.subtitle[
## Practical Lecture 4<br>Data Visualisation (base)
]
.author[
### Dr Louis Aslett
]
.institute[
### Durham University
]
.date[
### 31 October 2024
]

---

## Plotting data

- R has a wide array of tools for visualising data.

- So-called "exploratory data analysis" is really important, and can allow you to understand a data set a lot better.

- Never dive in to complicated modelling without first *looking* at your data!

- There are a few plotting frameworks in R, two most popular:
    - Base R
    - ggplot2

- ggplot2 can be made to produce publication quality graphics and is extremely powerful! However, it is a good bit more complicated than base R.

**DEMO:** motivation for looking at the data!

---
class: middle

## Presentation or exploration?

---

## Base R

Core plotting function:

``` r
plot(x, y, ...)
```

Provide vectors for `x` and `y`, which are points to plot.

Example of common extra arguments `...`:

- `col` to set colour of points (can use RGB or colour name as a string; can be vector for each point)
- `pch` change the plotting symbol (an integer, [click here for symbol numbers](https://cpb-us-e1.wpmucdn.com/sites.ucsc.edu/dist/d/276/files/2015/10/symbols.jpg))
- `xlab`, `ylab` x-axis or y-axis label
- `xlim`, `ylim` change the plotting range of x or y (by default uses data range)
- `main` plot title
- `type`
    - `"p"` default, points
    - `"l"` connect observations in order by lines
    - `"b"` points and lines simultaneously

If you only supply `x` then it will plot the values in `x` on the y-axis, against their index position in the vector on the x-axis.

---

## Base R

Each call to `plot()` makes a brand new plot. To add more points/lines (eg varying colour) to an existing plot:

``` r
points(x, y, ...)
lines(x, y, ...) # shorthand for points(x, y, type = "l")
```

- Similar options for `...`
- Some functions can be passed straight to lines:
    - `lowess()` to fit a smoothed line (`f` argument controls smoothness)
    - `density()` to fit a smoothed continuous version of histogram

Other plotting functions include:

- `hist()` for histograms
- `boxplot()` for boxplots
- `barplot()` for categorical bar charts (tip: use `table` to get summary first!)
- `abline()` add `$y=mx+c$` type straight lines to existing plot
    - can pass `lm()` directly to fit a straight line

All have slightly different interface, so part of learning R is learning to read the documentation!

---

## Breakneck tour! (I)

``` r
data("diamonds", package = "ggplot2")

plot(diamonds$carat, diamonds$price)

plot(diamonds$carat, diamonds$price, pch = 20)

plot(diamonds$carat, diamonds$price,
     pch = 20, xlim = c(0,1), ylim = c(0,10000))

plot(diamonds$carat, diamonds$price, pch = 20,
     main = "Dollar price against carat",
     xlab = "Carats", ylab = "Price in $")

abline(lm(price ~ carat, diamonds),
       col = "red")

lines(lowess(diamonds$carat, diamonds$price, f = 0.05),
      col = "green")
```

---

## Breakneck tour! (II)

``` r
hist(diamonds$price)

hist(diamonds$price, freq = FALSE)

hgram <- hist(diamonds$price, freq = FALSE)
hgram$breaks

hist(diamonds$price, freq = FALSE,
     breaks = seq(0, 20000, 2000))

mu.hat <- mean(diamonds$price)
sig.hat <- sd(diamonds$price)
x <- seq(0, 20000, length.out = 300)
lines(x, dnorm(x, mu.hat, sig.hat))

lines(density(diamonds$price), col = "red")

boxplot(diamonds$carat)

boxplot(diamonds$carat ~ diamonds$clarity)

cuts <- table(diamonds$cut)
barplot(cuts)
```

---

## Multiple plots

To get a grid of all pairwise scatter plots, use `pairs()`

``` r
pairs(mtcars)

pairs(mtcars[,1:4])
```

To manually create multiple plots, define the layout in terms of rows and columns

``` r
par(mfrow = c(2,1))
plot(diamonds$carat, diamonds$price)
boxplot(diamonds$carat)
par(mfrow = c(1,1)) # <- need this to reset to a single plot!
```

We will see better options with more flexible layouts soon!