```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE)
```

## Statistical Graphic?
![Stehpen Curry](figures/curry.png)

## Statistical Graphic?

![Stephen Curry-MVP](figures/curry_value.png)

## Statistical Graphic?

![Snow](figures/Snow-cholera-map-1.jpg)
- John Snow's Cholera map in dot style; dots represent deaths from cholera in London in 1854 to detect the source of the disease

## Statistical Graphic?

![Minard](figures/Minard1Vector.png)
- Charles Joseph Minard (1869), *Napoleon's March to Moscow - The War of 1812*


## Important questions for statistical graphics
- What is a graphic?
- How can we succinctly describe a graphic?
- How can we create the graphic that we have described?

One approach: develop a grammar!

- Grammar: *the fundamental principles
or rules of an art or science* (Oxford English Dictionary; Item 6)
    * Allows us to gain insights into the composition of complicated graphics
    * Reveals unexpected connections for understanding a diverse range of graphics
    * Guides us to produce sensical and well-formed graphics
- Analogy to the English language: good grammar is just the first step in creating a good sentence.

## Existing R graphics tools
- base graphics (Ross Ihaka)
    * pen on paper model; cannot modify or delete existing content
    * no representation of the graphics, apart from their appearance on the screen
    * fast but with limited scope
- `grid` (Paul Murrell, 2000) 
    * a much richer system of graphical primitives (only primitives; no tools for producing statistical graphics)
    * graph objects represented independently of the plot and can be modified later
    * a system of viewports to lay out complex graphics
- `lattice` package (Deepayan Sarkar, 2008)
    * uses `grid` to implement the trellis graphics system of Cleveland
    * can easily produce conditioned plots and some details (e.g., legends) are automatically taken care of
    * lacks a formal model; hard to extend
    
## `ggplot2`: a framework for producing statistical graphics
- takes the good things from base and lattice graphics
- uses a strong underlying model with several principles (details to follow)

What we get:

- a compact syntax to describe a wide range of graphics
- independent components that are easily extensible

## `ggplot2` Scatterplot Example: **data**

```{r basic1, echo=TRUE, message=FALSE, warning=FALSE,fig.width=12, fig.height=3,out.width='\\linewidth'}
# create a simple data set with 4 variables in the columns:
dat0 <- as.data.frame(list(A = c(2,1,4,9),
                          B = c(3,2,5,10),
                          C = c(4,1,15,80),
                          D = c("a","a","b","b")))
dat0
```

## `ggplot2` Scatterplot Example: *geom* aesthetics and **mapping**

- Scatterplot: 
    * a point for each observation
    * position the point horizontally according to the value of A, vertically according to C
    * Here, we will also map categorical variable D to the shape of the points

- Aesthetics:
    * x-poistion: A
    * y-position: C
    * shape: D

```{r basic2, echo=FALSE, message=FALSE, warning=FALSE,fig.width=12, fig.height=3,out.width='\\linewidth'}
# create a simple data set with 4 variables in the columns:
dat <- as.data.frame(list(x = c(2,1,4,9),
                          y = c(4,1,15,80),
                      Shape = c("a","a","b","b")))
print(dat)
```
## Example: mapping from data space to aesthetic space (controlled by **scale**)

```{r basic3, echo=TRUE, message=FALSE, warning=FALSE,fig.width=12, fig.height=3,out.width='\\linewidth'}
# ggplot2 creates (internally) a simple data set with 
# 3 variables in the columns:
dat_internal <- as.data.frame(list(x = c(25,0,75,200),
                          y = c(11,20,53,300),
                      Shape = c("circle","circle","square","square")))
print(dat_internal)

```


## Example: Plot the Geometric objects (**geom**)

```{r basic4, echo=TRUE, message=FALSE, warning=FALSE,fig.width=12, fig.height=3,out.width='\\linewidth'}
library(ggplot2)
example <- ggplot(data    = dat0,
                  mapping = aes(x=A,y=C,shape=D))+
           # Set aesthetics to fixed value
           geom_point(size = 7)
example

# run `?geom_point` geom_point understands the 
# following aesthetics (required aesthetics are in bold).

```

## Details about geometric objects, or **geom**
- Controls the type of the plot you create (a point **geom** creates a scatterplot; a line **geom** creates a line plot, etc.)
    * 0d: point, text,
    * 1d: path, line (ordered path),
    * 2d: polygon, interval.
- Are abstract and can be rendered in different ways (e.g., intervals).
- Require outputs from a statistic (e.g., x,y-positions in scatterplot; edges in boxplots)
- **Every *geom* has a default *stat*istic, and every *stat*istic a default *geom*.**
    * For example, the bin **stat**istic defaults to using the bar **geom** to produce a histogram. 
- Each **geom** can only display certain aesthetics. 
    * Try `?geom_point`
    * Different parameterizations may be useful (e.g., polar coordinate system).

## Example: Faceting (**facet**)

```{r basi5, echo=TRUE, message=FALSE, warning=FALSE,fig.width=12, fig.height=3,out.width='\\linewidth'}
library(ggplot2)
example <- ggplot(data    = dat0,
                  mapping = aes(x=A,y=C,shape=D))+
           # Set aesthetics to fixed value
           geom_point(size = 7)+facet_grid(~D)
example

```


## What are the components in the previous example? layered grammer of graphics (Wickham, 2009)
- **data** and **mappings** (describe how variables in the data are mapped to aesthetic attributes that you can perceive)
- geometric objects (**geom**); what you actually see on the plot, e.g., points, lines, polygons, etc.
- statistical transformations, **stat**; summarize data in useful ways, e.g., binning and counting to create a histogram
- **scale**: maps values in the data space to values in an aesthetic space; **scale** draws a legend or axes to make it possible to read the original data values from the graph (**inverse mapping**: what does this mean?)
- A coordinate system: **coord**
- A **facet**ing specification: describes how to break up data into subsets and how to display them as small multiples; also known as *conditioning* or *latticing/trelissing*.

## Diamond data

```{r diamond00, echo=TRUE, message=FALSE, warning=FALSE,fig.width=10, fig.height=6,out.width='\\linewidth'}
# just getting some data
library(ggplot2)
head(diamonds)
```

## Diamond data plotted by base graphics

```{r diamond01, echo=TRUE, message=FALSE, warning=FALSE,fig.width=10, fig.height=6,out.width='\\linewidth'}
plot(diamonds$carat, diamonds$price, col = diamonds$color,
     pch = as.numeric(diamonds$cut))
```

## Diamond data plotted by `ggplot2`
```{r diamond02, echo=TRUE, message=FALSE, warning=FALSE,fig.width=10, fig.height=6,out.width='\\linewidth'}
ggplot(diamonds, aes(carat, price, col = color, 
                     shape = cut)) +
    geom_point()
```

## Diamond Example: count within each cut category
```{r diamond1, echo=TRUE, message=FALSE, warning=FALSE,fig.width=10, fig.height=6,out.width='\\linewidth'}
d <- ggplot(diamonds, aes(cut))
d + geom_bar()
```

## Diamond Example: average prices within each cut category
```{r diamond2, echo=TRUE, message=FALSE, warning=FALSE,fig.width=12, fig.height=5,out.width='\\linewidth'}
d + stat_summary_bin(aes(y = price), fun.y = "mean", 
                     geom = "bar")
```

    
## Example: Napolean's March to Moscow

```{r minard1, echo=TRUE, message=FALSE, warning=FALSE,fig.width=12, fig.height=3,out.width='\\linewidth'}
library(HistData)
head(Minard.troops)
head(Minard.cities)
```

## Example: Napolean's March to Moscow

```{r minard2, echo=TRUE, message=FALSE, warning=FALSE,fig.width=12, fig.height=3,out.width='\\linewidth'}
data(Minard.troops); data(Minard.cities)
library(ggplot2)
plot_troops <- ggplot(Minard.troops, aes(long, lat)) +
  geom_path(aes(size = survivors, 
                colour = direction, group = group))

plot_troops

```

## Example: Napolean's March to Moscow

```{r minard3, echo=TRUE, message=FALSE, warning=FALSE,fig.width=12, fig.height=3,out.width='\\linewidth'}
plot_both <- plot_troops + 
  geom_text(aes(label = city), size = 4, 
            data = Minard.cities)
plot_both

```

## Example: Napolean's March to Moscow

```{r minard4, echo=TRUE, size="small", message=FALSE, warning=FALSE,fig.width=12, fig.height=3,out.width='\\linewidth'}

plot_polished <- plot_both + 
  scale_size(range = c(1, 12), 
             breaks = c(1, 2, 3) * 10^5, 
             labels = scales::comma(c(1, 2, 3) * 10^5)) + 
  scale_colour_manual(values = c("grey50","red")) +
  xlab(NULL) + 
  ylab(NULL)

plot_polished

```

## What grammar of graphics doesn't do

- It doesn't suggest what graphics you should use to answer the questions you are interested in. 
    * `ggplot2` focuses on how to produce the plots you want, not knowing what plots to produce.

- Grammar doesn't specify what a graphic should look like and how to make a plot attractive. 
    * Finer details, e.g., font size, background color are not specified by the grammar.
    * `ggplot2` uses its theming system
    
- No real-time interaction; other dynamic and interactive graphics packages exist: 
    * [rCharts: http://rcharts.io/](http://rcharts.io/)
    * [clickme: https://github.com/nachocab/clickme](https://github.com/nachocab/clickme)
    * [D3: Data-Driven Documents: https://d3js.org/](https://d3js.org/)

## "Ins and Outs"

- Data manipulation (get your data into the form required by `ggplot2`). You will shortly encounter at least these two R packages written by the same author of `ggplot2`:
    * `reshape2`
    * `plyr`
- Make figures publishable
    * comprehensive theming system in `ggplot2`

## References

- [Wickham H(2009). A Layered Grammar of Graphics. *Journal of Computational and Graphical Statistics*](http://byrneslab.net/classes/biol607/readings/wickham_layered-grammar.pdf)

Optional:

- Examples of statistical graphics used in sport analytics:
    * [Stephen Curry's Bombs Are Too Good To Be True - I mean, they have to be, right? (FiveThirtyEight.com, 2015)](http://fivethirtyeight.com/features/stephen-currys-bombs-are-too-good-to-be-true/)
    * [Lionel Messi Is Impossible (FiveThirtyEight.com, 2014)](http://fivethirtyeight.com/features/lionel-messi-is-impossible/)

- Notes by Wickham himself:
    * [`ggplot2` short courses by Wickham: http://courses.had.co.nz/11-rice/](http://courses.had.co.nz/11-rice/)
    * [`ggplot2` cheatsheet](https://www.rstudio.com/wp-content/uploads/2015/03/ggplot2-cheatsheet.pdf)

- A book-length introduction:
    * [Wickham (2010) `ggplot2`: *Elegant Graphics for Data Analysis (Use R!)* ](http://www.amazon.com/ggplot2-Elegant-Graphics-Data-Analysis/dp/0387981403)