---
title: "Introduction to ggplot2"
author: "Abhijit Dasgupta"
date: "BIOF 339"
---


```{r setup, include=F, child = here::here('slides/templates/setup.Rmd')}
```


```{r, echo=FALSE, results = 'asis'}
update_header()
```

---
class: middle, center

# Data visualization with ggplot2


---
## What is ggplot2?

- A second (and final) iteration of the ggplot
- Implementation of Wilkerson's Grammar of Graphics in R
- Conceptually, a way to layer different elements onto a canvas to create a data visualization
- Started as Dr. Hadley Wickham's PhD thesis (with Dr. Dianne Cook)
- Won the John M. Chambers Statistical Software Award in 2006

- Mimicked in other software platforms
  - `ggplot` and `seaborn` in Python
  - Translated in `plotly`

---

## ggplot2 uses the __grammar__ of __graphics__

.pull-left[
### A grammar ...

- compose and re-use small parts
- build complex structures from simpler units

]

--

.pull-right[
### of graphics ...

- Think of yourself as a painter
- Build a visualization  using layers on a canvas
- Draw layers on top of each other
]

---

## Introduction to ggplot2

The `ggplot2` package is a very flexible and (to me) intuitive way of visualizing data.
It is based on the concept of layering elements on a canvas.

> This idea of layering graphics on a canvas is, to me, a nice way of building graphs

You need:

+ A `data.frame` object
+ _Aesthetic mappings_ (aes) to say what data is used for what purpose in the viz
    + x- and y-direction
    + shapes, colors, lines
+ A _geometry object_ (geom) to say what to draw
    + You can "layer" geoms on each other to build plots
    
---
background-image: url(../img/grammar-of-graphics.png)
background-size: 80%,100%

---



## A dataset


```{r 05-Plotting-16, echo = T, message = F}
library(tidyverse)
library(rio)
beaches <- import('../data/sydneybeaches3.csv')
```
```{r 05-Plotting-17, echo = F}
head(beaches)
```

<p align="right" style="font-size: 10pt;">Credit: D. J. Navarro</p>

.footnote[Download the `sydneybeaches3.csv` file from the website and save it in your project's data folder]


---
class: middle, center

# Building a graph

---

.pull-left[
### Start with a blank canvas

```{r gr2, eval=F}
ggplot()
```
]
.pull-right[

### 

```{r, ref.label='gr2', eval=T, echo = F}

```

]

---

.pull-left[
### Add a data set

```{r gr3, eval=F}
ggplot(
  data = beaches # Tell ggplot the data you're using #<<
)
```

> Nothing has really happened yet, since we haven't said what we want to plot from the data set

> The `#` symbol tells R that anything after it is a *comment* and should be ignored. We'll make a lot of use of comments in our code.
]
.pull-right[

### 

```{r, ref.label='gr3', eval=T, echo = F}

```


]


---

.pull-left[
### Add a mapping from data to elements

```{r gr4, eval=F}
ggplot(
  data = beaches,
  mapping = aes( #<<
    x = temperature, #<<
    y = rainfall #<<
  )
)
```

What goes in 

- the x and y axes
- the color of markers
- the shape of markers
]
.pull-right[

### 

```{r, ref.label='gr4', eval=T, echo = F}

```

> Now we've said what we want to plot, so axes get drawn. We haven't yet specified how we want to plot it

]

---

.pull-left[
### Add a geometry to draw

```{r gr5, eval=F}
ggplot(
  data = beaches,
  mapping = aes(
    x = temperature,
    y = rainfall
  )
) +
  geom_point() #<<
```

What to draw:

- Points, lines
- histogram, bars, pies
]
.pull-right[

### 

```{r, ref.label='gr5', eval=T, echo = F}

```

> Now we've said **how** we want to plot the things we want from the data
]

---

.pull-left[
### Add options for the geom

```{r gr6, eval=F}
ggplot(
  data = beaches,
  mapping = aes(
    x = temperature,
    y = rainfall
  )
) +
  geom_point(size = 4) #<<
```

> Options pertaining to how we want things drawn are usually put in the function for the geometry, e.g. `geom_point`, `geom_line`, etc. If the option is based on any element from the dataset, we have to wrap it in a `aes()` function. We'll see this later
]
.pull-right[

### 

```{r, ref.label='gr6', eval=T, echo = F}

```

]

---

.pull-left[
### Add a mapping to modify the geom

```{r gr7, eval=F}
ggplot(
  data = beaches,
  mapping = aes(
    x = temperature,
    y = rainfall
  )
) +
  geom_point(
    mapping = aes(color = season_name), #<<
    size = 4
  )
```

> Anything data-driven has to be a mapping, driven by the `aes` function

]
.pull-right[

### 

```{r, ref.label='gr7', eval=T, echo = F}

```

]

---
class:middle, center

# A side note on `aes`

---

## The `aes` function

The `aes` function and when it needs to be used creates quite a bit of confusion initially. Let's start with the documentation for this function

```{r, eval=F}
?aes
```

> #### Description
Aesthetic mappings describe how variables in the data are mapped to visual properties (aesthetics) of geoms. Aesthetic mappings can be set in ggplot2() and in individual layers.

So, anytime we want to represent some aspect of the data on the plot in some form (color, shape, etc.), we have to use the `aes` function to *map* the data to the plot
---

## The `aes` function

The `aes` function can occur in one of two places:

.pull-left[
Within the `ggplot` function:

```{r, fig.height=2}
ggplot(
  data=beaches,
  mapping=aes(
    x = temperature,
    y = rainfall
  )
) + 
  geom_point()
```

]
.pull-right[
In the actual `geom` layer:

```{r, fig.height=2}
ggplot(
  data = beaches
) +
  geom_point(
    mapping = aes(
      x = temperature,
      y = rainfall
    )
  )
```

]

---

## The `aes` function

The `aes` function can occur in one of two places:

.pull-left[
Within the `ggplot` function:

+ You do this if the same mapping will be common to **all** the subsequent geometry layers
]
.pull-right[
In the actual `geom` layer:

+ You do this if the mapping will apply **only** to that layer
+ You do it if it makes more sense to put it in the `geom`. 
]
---
class: middle, center

# Facets / Trellis / Small multiples

---
.pull-left[
### Split into facets

```{r gr8, eval=F}
ggplot(
  data = beaches,
  mapping = aes(
    x = temperature,
    y = rainfall
  )
) +
  geom_point(
    mapping = aes(color = season_name),
    size = 4
  ) +
  facet_wrap( ~ season_name) #<<
```

> Create separate plots based on unique values of some variable (`season_name`) in your dataset

> Typically this variable should a few distinct values
]
.pull-right[

### 

```{r, ref.label='gr8', eval=T, echo = F}

```

]

---

.pull-left[
### Remove the legend

```{r gr9, eval=F}
ggplot(
  data = beaches,
  mapping = aes(
    x = temperature,
    y = rainfall
  )
) +
  geom_point(
    mapping = aes(color = season_name),
    size = 4,
    show.legend = FALSE #<<
  ) +
  facet_wrap( ~ season_name) 
```

> Now we're getting more into the look of the plot and how much information should be on it

> `show.legend` option is in `geom_point` here since it would be based on the colors of the points. If you had a different geometry like `geom_line`, you would put the `show.legend` option there if the legend was based on that geom.
]
.pull-right[

### 

```{r, ref.label='gr9', eval=T, echo = F}

```

]

---

.pull-left[
### Change the background

```{r gr10, eval=F}
ggplot(
  data = beaches,
  mapping = aes(
    x = temperature,
    y = rainfall
  )
) +
  geom_point(
    mapping = aes(color = season_name),
    size = 4,
    show.legend = FALSE
  ) +
  facet_wrap( ~ season_name) +
  theme_bw() #<<
```

> Again, a look-and-feel choice

> In-built `ggplot` themes are described [here](https://ggplot2.tidyverse.org/reference/ggtheme.html). Other themes are availabe via packages [`ggthemes`](https://yutannihilation.github.io/allYourFigureAreBelongToUs/ggthemes/) and others.

]
.pull-right[

### 

```{r, ref.label='gr10', eval=T, echo = F}

```

]

---

.pull-left[
### Update the labels

```{r gr11, eval=F}
ggplot(
  data = beaches,
  mapping = aes(
    x = temperature,
    y = rainfall
  )
) +
  geom_point(
    mapping = aes(color = season_name),
    size = 4,
    show.legend = FALSE
  ) +
  facet_wrap( ~ season_name) +
  theme_bw() +
  labs(x = 'Temperature (C)', y = 'Rainfall (mm)') #<<
```

> This is **important**. Make sure the information in your plot is self-contained by putting appropriate labels and titles on it.
]
.pull-right[

### 

```{r, ref.label='gr11', eval=T, echo = F}

```

]

---

.pull-left[
### Add titles

```{r gr12, eval=F}
ggplot(
  data = beaches,
  mapping = aes(
    x = temperature,
    y = rainfall
  )
) +
  geom_point(
    mapping = aes(color = season_name),
    size = 4,
    show.legend = FALSE
  ) +
  facet_wrap( ~ season_name) +
  theme_bw() +
  labs(x = 'Temperature (C)', 
       y = 'Rainfall (mm)',
       title = 'Sydney weather by season', #<<
       subtitle = "Data from 2013 to 2018") #<<
```

]
.pull-right[

### 

```{r, ref.label='gr12', eval=T, echo = F}

```

]

---

.pull-left[
### Customize

```{r gr13, eval=F}
ggplot(
  data = beaches,
  mapping = aes(
    x = temperature,
    y = rainfall
  )
) +
  geom_point(
    mapping = aes(color = season_name),
    size = 4,
    show.legend = FALSE
  ) +
  facet_wrap( ~ season_name) +
  theme_bw() +
  labs(x = 'Temperature (C)', 
       y = 'Rainfall (mm)',
       title = 'Sydney weather by season', 
       subtitle = "Data from 2013 to 2018") +
  theme(axis.title = element_text(size = 14), #<<
        axis.text = element_text(size = 12), #<<
        strip.text = element_text(size = 12)) #<<
```

]
.pull-right[

### 

```{r, ref.label='gr13', eval=T, echo = F}

```

> Once again, a look-and-feel choice, but this is very useful for publications when the actual figure will be smaller but we need the labels to be legible.
]

---
class: middle, center

# Understanding the structure of ggplot

---
background-image: url(../img/grammar.png)
background-size: contain

---

## The grammar
.left-column70[
- Data
- Aesthetics (or aesthetic mappings)
- Geometries (as layers)
- Facets
- Themes
- (Coordinates)
- (Scales)
]
.right-column70[
> Data, Aesthetics and Geometries are **required** to actually create a plot
]

---
class: middle, center

# Peeking under the hood

---
.pull-left[
### If I write...

```{r, eval = F, echo = T}
ggplot(
  data = beaches,
  aes(x = temperature,
      y = rainfall)
) + 
  geom_point()
```

]
--
.pull-right[
### what's really run is ...

```{r,echo = T, eval = F}
ggplot(
  data = beaches, 
  mapping = aes(
    x = temperature, y = rainfall)) + 
layer(
  geom = "point",
  stat = "identity",
  position = "identity") + 
facet_null() +
theme_grey() + 
coord_cartesian() + 
scale_x_continuous() + 
scale_y_continuous()
```

]

--

### Each element can be adapted and tweaked to create graphs

---
class: middle, center

# Exploring aesthetics, mappings and their uses

---

## Let's start

.pull-left[
```{r grp1, eval = F, echo = T}
ggplot(
  data = beaches, 
  mapping = aes(
    x = enterococci
  )
) + 
  geom_density()
```

> This is looking at the distribution of enterococci concentration (x-axis) in this dataset
]
.pull-right[
```{r, eval=T, echo = F, ref.label="grp1"}
```
]

---

## Bacteria growth should depend on temperature....

.pull-left[
```{r grp2, eval = F, echo = T}
ggplot(
  data = beaches,
  mapping = aes(
    x = enterococci,
    group = season_name #<<
  )
) + 
  geom_density()
```

> Can't really parse the seasons out from the graph. 

]
.pull-right[
```{r, eval=T, echo = F, ref.label="grp2"}
```
]

---

## Add some color

.pull-left[
```{r grp3, eval = F, echo = T}
ggplot(
  data = beaches,
  mapping = aes(
    x = enterococci, 
    group = season_name,
    color = season_name #<<
  )
) + 
  geom_density()
```

> This does better distinguish the seasons (and we get a legend as a by-product). 

> But.... things are too crammed on the left of the plot. This is because the bacterial concentrations are low, with some high values. A transformation may be nice.
]
.pull-right[
```{r, eval=T, echo = F, ref.label="grp3"}
```
]

---

## A better choice of scale

.pull-left[
```{r grp5, eval = F, echo = T}
ggplot(
  data = beaches,
  mapping = aes(
    x = enterococci,
    group = season_name,
    color = season_name
  )
) + 
  geom_density() + 
  scale_x_log10() #<<
```

> This makes things a bit clearer, but still a bunch of squiggly lines
]
.pull-right[
```{r, eval=T, echo = F, ref.label="grp5"}
```
]

---

## A better choice of where to put the color

.pull-left[
```{r grp6, eval = F, echo = T}
ggplot(
  data = beaches, 
  mapping = aes(
    x = enterococci,
    group = season_name,
    fill = season_name #<<
  )
) +
  geom_density() + 
  scale_x_log10()
```

> Things are covered up!

> For geometries with "insides", `color` puts color on the outlines, and `fill` puts color on the insides.
]
.pull-right[
```{r, eval=T, echo = F, ref.label="grp6"}
```
]

---

## Let's play peek-a-boo

.pull-left[
```{r grp7, eval = F, echo = T}
ggplot(
  data = beaches, 
  mapping = aes(
    x = enterococci,
    group = season_name,
    fill = season_name #<<
  )
) +
  geom_density(alpha = 0.3) + #<< 
  scale_x_log10()

```

> A bit better, but not great

]
.pull-right[
```{r, eval=T, echo = F, ref.label="grp7"}
```
]

---

## Let's break things out

.pull-left[
```{r grp8, eval = F, echo = T}
ggplot(
  data = beaches, 
  mapping = aes(
    x = enterococci,
    group = season_name,
    fill = season_name #<<
  )
) +
  geom_density(alpha = 0.3) + 
  scale_x_log10() + 
  facet_wrap(~ season_name) #<<

```

> A lot clearer!!

> There's stuff we don't need now, like the legend
> Also, can we please make the numbers human-readable
]
.pull-right[
```{r, eval=T, echo = F, ref.label="grp8"}
```
]

---

## Cleaning up

.pull-left[
```{r grp10, eval = F, echo = T}
ggplot(
  data = beaches, 
  mapping = aes(
    x = enterococci,
    fill = season_name
  )
) +
  geom_density(show.legend = FALSE) + 
  scale_x_log10(labels = scales::label_number()) + 
  facet_wrap(~season_name) 
    

```
]
.pull-right[
```{r, eval=T, echo = F, ref.label="grp10"}
```
]
---


## Fixing the legend ordering

.pull-left[
```{r gr17, eval = F}
beaches <- beaches %>%  #<<
  mutate(season_name = fct_relevel(season_name,#<<
                                   c('Spring','Summer','Autumn','Winter')))#<<
ggplot(
  data=beaches,
  aes(x = enterococci,
      fill = season_name#<<
  )
) + 
  geom_density(show.legend = F) + 
  facet_wrap(~ season_name) + 
  scale_x_log10(labels = scales::label_number())
```

]
.pull-right[
```{r, ref.label='gr17', eval=T, echo=F}

```

]


---
class: center, middle

# Exploring geometries

---
class: center, middle

# Univariate plots

---

## Histograms

.pull-left[
```{r gr20, eval = F}
library(tidyverse)
library(rio)
dat_spine <- import('../data/Dataset_spine.csv', 
                    check.names = T) %>% 
  janitor::clean_names()

ggplot(
  data=dat_spine,
  aes(x = degree_spondylolisthesis))+
  geom_histogram()
```
]
.pull-right[
```{r, ref.label='gr20', echo = F, eval=T, message = T}

```

]

---


## Histograms

.pull-left[
```{r gr21, eval = F}

ggplot(
  data=dat_spine,
  aes(x = degree_spondylolisthesis))+
  geom_histogram(bins = 100) #<<
```

This gives a very different view of the data
]
.pull-right[
```{r, ref.label='gr21', echo = F, eval=T, message = T}

```

]

---

## Density plots

.pull-left[
```{r gr22, eval = F}
ggplot(
  data=dat_spine,
  aes(x = degree_spondylolisthesis))+
  geom_density() #<<
```
]
.pull-right[
```{r, ref.label='gr22', echo = F, eval=T, message = T}

```

]

---

## Density plots

.pull-left[
```{r gr23, eval = F}
ggplot(
  data=dat_spine,
  aes(x = degree_spondylolisthesis))+
  geom_density(adjust = 1/5) # Use 1/5 the bandwidth #<<
```
]
.pull-right[
```{r, ref.label='gr23', echo = F, eval=T, message = T}

```

]

---

## Layering geometries

.pull-left[
```{r gr24, eval = F}
ggplot(
  data=dat_spine,
  aes(x = degree_spondylolisthesis,
      y = stat(density)))+  # Re-scales histogram #<<
  geom_histogram(bins = 100, fill='yellow') +
  geom_density(adjust = 1/5, color = 'orange', size = 2)
```

]
.pull-right[
```{r, ref.label='gr24', eval=T, echo=F}

```

]

---

## Bar plots (categorical variable)

.pull-left[
```{r gr25, eval=F, echo=T}
dat_brca <- rio::import('../data/clinical_data_breast_cancer_modified.csv', 
                   check.names = T)
ggplot(
  data=dat_brca,
  aes(x = Tumor))+
  geom_bar()
```
]
.pull-right[
```{r,ref.label='gr25', eval=T, echo = F}

```

]

---

## Bar plots (categorical variable)

.pull-left[
```{r gr26, eval=F, echo=T}
dat_brca <- import('../data/clinical_data_breast_cancer_modified.csv', 
                   check.names = T)
ggplot(
  data=dat_brca,
  aes(x = Tumor,
      fill = ER.Status))+ #<<
  geom_bar()
```

Add additional information via mapping
]
.pull-right[
```{r,ref.label='gr26', eval=T, echo = F}

```
]

---

## Bar plots (categorical variable)

.pull-left[
```{r gr27, eval=F, echo=T}
dat_brca <- import('../data/clinical_data_breast_cancer_modified.csv', 
                   check.names = T)
ggplot(
  data=dat_brca,
  aes(x = Tumor,
      fill = ER.Status))+ 
  geom_bar(position = 'dodge')  #<<
    # Default is position = "stack"
```

Change the nature of the geometry
]
.pull-right[
```{r,ref.label='gr27', eval=T, echo = F}

```
]

???

It is not easy to do patterns in R. Its been considered a bug rather than a feature,
since you can induce cognitive problems using patterns. It is suggested to use 
different palettes for distinguishing groups

---

## Graphing tabulated data

.pull-left[
```{r, eval = T, warning = T, error = T}
tabulated <- dat_brca %>% count(Tumor)
tabulated

ggplot(
  data = tabulated,
  aes(x = Tumor, y = n)) +
  geom_bar()
```
]

---

## Graphing tabulated data

.pull-left[
```{r gr_identity, eval = F, echo = T, warning = T}
tabulated <- dat_brca %>% count(Tumor)
tabulated

ggplot(
  data = tabulated,
  aes(x = Tumor, y = n)) +
  geom_bar(stat = 'identity') #<<
```

Here we need to change the default computation    

The barplot usually computes the counts (`stat_count`)  

We suppress that here since we have already   
done the computation
]
.pull-right[
```{r, ref.label='gr_identity', eval = T, results='hide',echo = F}

```

]

---

## Peeking under the hood

.pull-left[
```{r}
plt <- ggplot(
  data = tabulated,
  aes(x = Tumor, y = n)) +
  geom_bar()

plt$layers
```

]
.pull-right[
```{r}
plt <- ggplot(
  data = tabulated,
  aes(x = Tumor, y = n)) +
  geom_bar(stat = 'identity')

plt$layers
```

]

--

> Each layer has a geometry, statistic and position associated with it

---
class: middle, center

# Bivariate plots

---

## Scatter plots

.pull-left[
```{r gr28, eval=F}
ggplot(
  data = beaches,
  aes(x = date, y = temperature))+
  geom_point()

```

This is great, but the dates are being treated as individual levels, since
dates are encoded as characters `r fontawesome::fa('grimace', fill='red')`. 

```{r}
beaches <- beaches %>% 
  mutate(date1 = as.Date(date))
```

]
.pull-right[
```{r, ref.label='gr28', eval=T, echo = F}

```

]

---

## Scatter plots

.pull-left[
```{r gr281, eval=F}
ggplot(
  data = beaches,
  aes(x = date1, y = temperature))+
  geom_point()

```
]
.pull-right[
```{r, ref.label='gr281', eval=T, echo = F}

```

]

---

## Scatter plots

.pull-left[
```{r gr29, eval=F}
ggplot(
  data = beaches,
  aes(x = date1, y = temperature))+
  geom_line() #<<

```

> If you keep `date` as a character, you dont see any lines plotted, and get the following warning:
> 
> `geom_path: Each group consists of only one observation. Do you
need to adjust the group aesthetic?`
> 
> This is cryptic!! This arises since `date` is a character, gets converted to a `factor` for plotting, and then R doesn't know how to join lines between factors.
]
.pull-right[
```{r, ref.label='gr29', eval=T, echo = F}

```

]

---

## Scatter plots

.pull-left[
```{r gr30, eval=F}
ggplot(
  data = beaches,
  aes(x = date1, y = temperature))+
  geom_point(color='black', size = 3) +
  geom_line(color='red',size=2)

```

Layer points and lines
]
.pull-right[
```{r, ref.label='gr30', eval=T, echo = F}

```

]

---

## Scatter plots

.pull-left[
```{r gr31, eval=F}
ggplot(
  data = beaches,
  aes(x = date1, y = temperature))+
  geom_line(color='red',size=2) + #<<
  geom_point(color='black', size = 3) #<<

```

Order of laying down geometries matters

]
.pull-right[
```{r, ref.label='gr31', eval=T, echo = F}

```

]

---

## Doing some computations

.pull-left[
```{r gr32, eval = F, echo = T}
ggplot(
  data = beaches,
  aes(x = date1, y = temperature)) +
  geom_point() +
  geom_smooth() #<<
```

Averages over 75% of the data
]
.pull-right[
```{r, ref.label='gr32', eval=T, echo=F}

```

]

---

## Doing some computations

.pull-left[
```{r gr33, eval = F, echo = T}
ggplot(
  data = beaches,
  aes(x = date1, y = temperature)) +
  geom_point() +
  geom_smooth(span = 0.1) #<<
```

Averages over 10% of the data
]
.pull-right[
```{r, ref.label='gr33', eval=T, echo=F}

```

]

---

## Computations over groups

.pull-left[
```{r gr34, echo = T, eval = F}
ggplot(
  data = dat_spine,
  aes(x = sacral_slope, 
      y = degree_spondylolisthesis)) +
  geom_point() +
  geom_smooth()
```
]
.pull-right[
```{r,ref.label='gr34', echo = F, eval = T, message = F}

```

]

---

## Computations over groups

.pull-left[
```{r gr35, echo = T, eval = F}
ggplot(
  data = dat_spine,
  aes(x = sacral_slope, 
      y = degree_spondylolisthesis,
      color = class_attribute)) + #<<
  geom_point() +
  geom_smooth()
```

Computation is done by groups
]
.pull-right[
```{r,ref.label='gr35', echo = F, eval = T, message = F}

```

]

---

## Computations over groups

.pull-left[
```{r gr36, echo = T, eval = F}
ggplot(
  data = dat_spine,
  aes(x = sacral_slope, 
      y = degree_spondylolisthesis,
      color = class_attribute)) + 
  geom_point() +
  geom_smooth() +
  xlim(0, 100)#<<
```

Ignore the outlier for now
]
.pull-right[
```{r,ref.label='gr36', echo = F, eval = T, message = F}

```

]

---

## Computations over groups

.pull-left[
```{r gr37, echo = T, eval = F}
ggplot(
  data = dat_spine,
  aes(x = sacral_slope, 
      y = degree_spondylolisthesis)) +
  geom_point() +
  geom_smooth(aes(color = class_attribute)) + #<<
  xlim(0, 100)
```

Only color-code the smoothers   

You can change the plot based on where you map the aesthetic
]
.pull-right[
```{r,ref.label='gr37', echo = F, eval = T, message = F}

```

]

---

## Computations over groups

.pull-left[
```{r gr38, echo = T, eval = F}
ggplot(
  data = dat_spine,
  aes(x = sacral_slope, 
      y = degree_spondylolisthesis)) +
  geom_point() +
  geom_smooth(aes(color = class_attribute),
              se = F) + #<<
  xlim(0, 100)
```

Remove the confidence bands    

Maybe a cleaner look
]
.pull-right[
```{r,ref.label='gr38', echo = F, eval = T, message = F}

```

]


---

## Box Plots

.pull-left[
```{r gr39, echo = T, eval = F}
ggplot(
  data = beaches,
  aes(x = season_name, 
      y = temperature)) + 
  geom_boxplot()
```

]
.pull-right[
```{r, ref.label='gr39', eval = T, echo = F}

```
]

---

## Box Plots

.pull-left[
```{r, echo = T, eval = F}
ggplot(
  data = beaches,
  aes(x = season_name, 
      y = temperature)) + 
  geom_boxplot()
```

]
.pull-right[
```{r,  eval = T, echo = F}
p <- ggplot(
  data = beaches,
  aes(x = season_name, 
      y = temperature)) + 
  geom_boxplot()
p + annotate('text', x = 2, y = 18, label = 'Minimum') + 
  annotate('text', x = 2.5, y = 37, label = 'Maximum (within fence)') +
  annotate('text', x = 2.0, y = 30.3, label='3rd quartile', vjust = 1) +
  annotate('text', x = 2.0, y = 25.7, label = '1st quartile', vjust = 0) + 
  annotate('text', x = 2.0, y = 27.9, label = 'Median', vjust = 0) +
  annotate('text', x = 2.1, y = 40, label = '"Outliers"', angle = -90)

```
]

---

## Violin plots

.pull-left[
```{r gr40, echo = T, eval = F}
ggplot(
  data = beaches,
  aes(x = season_name, 
      y = temperature)) + 
  geom_violin() #<<
```

]
.pull-right[
```{r, ref.label='gr40', eval = T, echo = F}

```
]

---

## Strip plots

.pull-left[
```{r gr41, echo = T, eval = F}
ggplot(
  data = beaches,
  aes(x = season_name, 
      y = temperature)) + 
  geom_point()
```

Points are overlayed on each other
]
.pull-right[
```{r, ref.label='gr41', eval = T, echo = F}

```
]

---

## Strip plots

.pull-left[
```{r gr42, echo = T, eval = F}
ggplot(
  data = beaches,
  aes(x = season_name, 
      y = temperature)) + 
  geom_jitter() #<<
```

Jittering allows all points to be seen

Maybe too much
]
.pull-right[
```{r, ref.label='gr42', eval = T, echo = F}

```
]

---

## Strip plots

.pull-left[
```{r gr43, echo = T, eval = F}
ggplot(
  data = beaches,
  aes(x = season_name, 
      y = temperature)) + 
  geom_jitter(width = 0.2) #<<
```

Jittering allows all points to be seen

Maybe too much
]
.pull-right[
```{r, ref.label='gr43', eval = T, echo = F}

```
]

---

## Layers, again

.pull-left[
```{r gr44, echo = T, eval = F}
ggplot(
  data = beaches,
  aes(x = season_name, 
      y = temperature)) + 
  geom_boxplot() + #<<
  geom_jitter(width = 0.2)
```


]
.pull-right[
```{r, ref.label='gr44', eval = T, echo = F}

```
]

---

## Layers, again

.pull-left[
```{r gr45, echo = T, eval = F}
ggplot(
  data = beaches,
  aes(x = season_name, 
      y = temperature)) + 
  geom_violin() + #<<
  geom_jitter(width = 0.2)
```


]
.pull-right[
```{r, ref.label='gr45', eval = T, echo = F}

```
]




---
class: middle, center

# Scales

---

.pull-left[
```{r gr53, eval = F, echo =T}
ggplot(
  data = beaches,
  aes(x = date1, y = enterococci)) + 
  geom_point()
```

All the action is happening in the bottom bit
]
.pull-right[
```{r, ref.label='gr53', eval = T, echo = F}

```

]

---

.pull-left[
```{r gr54, eval = F, echo =T}
ggplot(
  data = beaches,
  aes(x = date1, y = enterococci)) + 
  geom_point() + 
  scale_y_log10() #<<
```

Log-transforming an axis can make things easier to see
]
.pull-right[
```{r, ref.label='gr54', eval = T, echo = F}

```

]

---

.pull-left[
```{r gr55, eval = F, echo =T}
ggplot(
  data = beaches,
  aes(x = date1, y = enterococci)) + 
  geom_point() + 
  scale_y_log10( #<<
    labels = scales::number_format(digits=3)) #<<
```

Making the labels a bit easier to read
]
.pull-right[
```{r, ref.label='gr55', eval = T, echo = F}

```

]

???

`scales::number_format` is calling the function `number_format` from the 
package `scales`

---
class: middle, center

# Order and orientation

---

# Arrests in the USA in 1973

.pull-left[
```{r gr56, eval = F, echo = T}
arrests <- import('../data/USArrests.csv')
ggplot(
  data = arrests,
  aes(x = State, 
      y = Murder)) + 
  geom_bar(stat = 'identity')
```

This plot is very hard to read

There is no ordering, and states can't be read
]
.pull-right[
```{r, ref.label='gr56', eval = T, echo = F}

```

]

---
# Arrests in the USA in 1973

.pull-left[
```{r gr561, eval = F, echo = T}
arrests <- import('../data/USArrests.csv')
ggplot(
  data = arrests,
  aes(x = fct_reorder(State, Murder), # Order by murder rate #<<
      y = Murder)) + 
  geom_bar(stat = 'identity')
```

We see the pattern, but its still unreadable
]
.pull-right[
```{r, ref.label='gr561', eval = T, echo = F}


```

]

---

# Arrests in the USA in 1973

.pull-left[
```{r gr57, eval = F, echo = T}
arrests <- import('../data/USArrests.csv')
ggplot(
  data = arrests,
  aes(x = fct_reorder(State, Murder), # Order by murder rate 
      y = Murder)) + 
  geom_bar(stat = 'identity') + 
  coord_flip() #<<
```

Flipping the axes makes the states readable

]
.pull-right[
```{r, ref.label='gr57', eval = T, echo = F}

```

]

---

# Arrests in the USA in 1973

.pull-left[
```{r gr58, eval = F, echo = T}
arrests <- import('../data/USArrests.csv')
ggplot(
  data = arrests,
  aes(x = fct_reorder(State, Murder), # Order by murder rate 
      y = Murder)) + 
  geom_bar(stat = 'identity') + 
  labs(x = 'State', y = 'Murder rate') + #<<
  theme_bw() +#<<
  theme(panel.grid.major.y = element_blank(),#<<
        panel.grid.minor.y = element_blank()) + #<<
  coord_flip()
```

Cleaning it up a little


]
.pull-right[
```{r, ref.label='gr58', eval = T, echo = F}

```

]

???

Note that we flip after we've done all the manipulation

---
class: middle, center

# Themes

---

# Color schemes

ggplot comes with a default color scheme. There are several other schemes available

- `scale_*_brewer` uses the [ColorBrewer](http://colorbrewer2.org) palettes
- `scale_*_gradient` uses gradients
- `scale_*_distill` uses the ColorBrewer palettes, for continuous outcomes

> Here * can be `color` or `fill`, depending on what you want to color
>
> Note `color` refers to the outline, and `fill` refers to the inside

---

.pull-left[
```{r gr60, eval = F, echo = T}
ggplot(
  data = beaches,
  aes(x = date, y = enterococci,
      color = temperature)) + 
  geom_point() +
  scale_y_log10(name = 'Enterococci',
                label = scales::number_format(digits=3))
```

]
.pull-right[
```{r, ref.label='gr60', eval = T, echo = F}

```
]
---

.pull-left[
```{r gr61, eval = F, echo = T}
ggplot(
  data = beaches,
  aes(x = date, y = enterococci,
      color = temperature)) + 
  geom_point() +
  scale_y_log10(name = 'Enterococci',
                label = scales::number_format(digits=3))+
  scale_color_gradient(low = 'white', high='red') + 
  theme_dark()
```

]
.pull-right[
```{r, ref.label='gr61', eval = T, echo = F}

```
]
---

.pull-left[
```{r gr62, eval = F, echo = T}
ggplot(
  data = beaches,
  aes(x = date, y = enterococci,
      color = temperature)) + 
  geom_point() +
  scale_y_log10(name = 'Enterococci',
                label = scales::number_format(digits=3))+
  scale_color_gradient(low = 'blue', high='red') + 
  theme_bw()
```

]
.pull-right[
```{r, ref.label='gr62', eval = T, echo = F}

```
]
---

## Specifying colors

.pull-left[
```{r gr63, echo = T, eval = F}
ggplot(
  data = dat_spine,
  aes(x = sacral_slope, y = degree_spondylolisthesis,
      color = class_attribute)) +
  geom_point() + 
  geom_smooth(se = F) + 
  coord_cartesian(xlim = c(0, 100), ylim = c(0,200))
```

]
.pull-right[
```{r, ref.label='gr63', eval = T, echo = F, message = F}

```

]

---

# Specifying colors

.pull-left[
```{r gr64, echo = T, eval = F}
ggplot(
  data = dat_spine,
  aes(x = sacral_slope, y = degree_spondylolisthesis,
      color = class_attribute)) +
  geom_point() + 
  geom_smooth(se = F) + 
  coord_cartesian(xlim = c(0, 100), ylim = c(0,200)) + 
  scale_color_manual(values = c("Normal"="blue", 'Abnormal' = 'red'))
```

]
.pull-right[
```{r, ref.label='gr64', eval = T, echo = F, message = F}

```

]

---

## A note on colors

When choosing colors, you should consider issues of accessibility, especially regarding color-blindness

.pull-left[
![](../img/cvd1.png)
]
.pull-right[
![](../img/cvd2.png)
]

---

## Themes

You can create your own custom themes to keep a unified look to your graphs

ggplot comes with

- theme_classic
- theme_bw
- theme_void
- theme_dark
- theme_gray
- theme_light
- theme_minimal

---

## Themes

.pull-left[
### Create your own

```{r gr65, echo = T, eval = F}

ggplot(
  data = dat_spine,
  aes(x = sacral_slope, y = degree_spondylolisthesis,
      color = class_attribute)) +
  geom_point() + 
  geom_smooth(se = F) + 
  coord_cartesian(xlim = c(0, 100), 
                  ylim = c(0,200))
```

]
.pull-right[
```{r, ref.label='gr65', eval = T, echo = F, message = F}

```
]

---

## Themes

.pull-left[
### Create your own

```{r gr66, echo = T, eval = F}

my_theme <- function(){
  theme_bw()
}

ggplot(
  data = dat_spine,
  aes(x = sacral_slope, y = degree_spondylolisthesis,
      color = class_attribute)) +
  geom_point() + 
  geom_smooth(se = F) + 
  coord_cartesian(xlim = c(0, 100), 
                  ylim = c(0,200)) +
  my_theme() #<<
```

]
.pull-right[
```{r, ref.label='gr66', eval = T, echo = F, message = F}

```
]


---

## Themes

.pull-left[
### Create your own

```{r gr68, echo = T, eval = F}

my_theme <- function(){
  theme_bw() +
  theme(axis.text = element_text(size = 14),
        axis.title = element_text(size = 16),
        panel.grid.minor = element_blank(),
        strip.text = element_text(size=14),
        strip.background = element_blank())
}

ggplot(
  data = dat_brca,
  aes(x = Tumor))+
  geom_bar() + 
  facet_grid(rows = vars(ER.Status),
             cols = vars(PR.Status))
```

]
.pull-right[
```{r, ref.label='gr68', eval = T, echo = F}

```
]

---

## Themes

.pull-left[
### Create your own

```{r gr69, echo = T, eval = F}

my_theme <- function(){
  theme_bw() +
  theme(axis.text = element_text(size = 14),
        axis.title = element_text(size = 16),
        panel.grid.minor = element_blank(),
        strip.text = element_text(size=14),
        strip.background = element_blank())
}

ggplot(
  data = dat_brca,
  aes(x = Tumor))+
  geom_bar() + 
  facet_grid(rows = vars(ER.Status),
             cols = vars(PR.Status)) +
  my_theme() #<<
```

]
.pull-right[
```{r, ref.label='gr69', eval = T, echo = F}

```
]