notes.Rmd

---
title: "Stat 33A - Lecture 4"
date: March 2, 2020
output: pdf_document
---

Announcements
=============

Quiz 2 this week! Same format as quiz 1.


Review
======

Last time:

* Paths
* Inspecting data frames

In the notebook, if you use `setwd()` it only lasts for __that
chunk__ and is then reset:


So in subsequent chunks it looks like you didn't call `setwd()`:
```{r}
getwd()

setwd("..")

getwd()
```

```{r}
getwd()
```


Plan ahead so that other people can run your code and reproduce your
results.

Good habits:

* Putting your notebook(s) and data in the project directory.
* Using paths relative to the project directory.

Bad habits:

* Calling `setwd()` in R notebooks and scripts.
* Using absolute paths.

It's okay to use `setwd()` in the *R console* to set the working
directory to your project directory.


R Packages
==========

The Comprehensive R Archive Network (CRAN) is a repository of
user-contributed packages for R.

You can install packages from CRAN with `install.packages()`:
```{r}
?install.packages
```

For maintaining your packages, there are also functions:

* `remove.packages()` to remove a package
* `update.packages()` to update ALL packages
* `installed.packages()` to list installed packages

```{r}
installed.packages()
```


You can load an installed package into your R session with the
`library()` function:
```{r}
library(dplyr)

library("dplyr")
```


Install once, load with `library()` each time you restart R.


Working with Data with `dplyr`
==============================

The `dplyr` package provides functions for working with data frames.

We'll use `dplyr` for now, and learn about R's built-in tools later.

Cheat sheet:

    https://github.com/rstudio/cheatsheets/raw/master/
        data-transformation.pdf

```{r}
dogs = readRDS("data/dogs/dogs_full.rds")

head(dogs)
```


Use `slice()` to choose rows by position:
```{r}
slice(dogs, 4)

slice(dogs, c(1, 3, 3, 5))

slice(dogs, 6:9)
```

Use `filter()` to choose rows that satisfy a condition:
```{r}
5 < 6

c(1, 3, 4) < c(5, 2, 1)

filter(dogs, group == "toy")
```
Suppose we want all toy dogs and all sporting dogs:
```{r}
filter(dogs, group == "toy")
filter(dogs, group == "sporting")

# Operators for combining conditions: &, AND
TRUE & TRUE

TRUE & FALSE

FALSE & TRUE

FALSE & FALSE

# Another option for combining: |, OR
TRUE | TRUE

TRUE | FALSE

FALSE | FALSE

# How the two groups of dogs?

filter(dogs, group == "toy" & group == "sporting") # no good

filter(dogs, group == "toy" | group == "sporting") # this works
```

Suppose we want to get all dogs with group "toy" AND with weight < 10:
```{r}
light_dogs = filter(dogs, group == "toy" & weight < 10)

# Check weight column:
light_dogs$weight

# Behind the scenes:
dogs$group == "toy"
```


Use `select()` to choose columns by name or position:
```{r}
class(select(dogs, c(weight, height)))

class(select(dogs, weight))

class(dogs$weight)

# Use vectors for arithmetic:
dogs$weight + dogs$height

# Rather than columns:
select(dogs, weight) + select(dogs, height) 
```

Use `:` to indicate a range of rows or columns:
```{r}
names(dogs)

select(dogs, 1:3)

# Only in dplyr can we use names of columns with `:`
select(dogs, price:kids)
```

Use `-` to exclude rows or columns:
```{r}
# Order of operations is important:
-1:3
-(1:3)

slice(dogs, -(1:3))

select(dogs, -1)

# Only for dplyr:
select(dogs, -datadog)
```


All dogs not in the toy group (how to invert a condition with `filter()`):
```{r}
filter(dogs, group != "toy")

# Another way, useful for long conditions:
filter(dogs, !(group == "toy"))

!c(FALSE, TRUE, TRUE)
```


Other useful `dplyr` functions:

* `arrange()` changes the ordering of the rows.
* `mutate()` adds new columns by transforming existing columns.
* `summarise()` reduces multiple rows down to a single value.
* `group_by()` splits rows into groups when summarizing.



The Tidyverse
=============

The Tidyverse (<https://www.tidyverse.org/>) is collection of R
packages for doing data science in R.

They are made by many of the same people that make RStudio, and
provide alternatives to R's built-in tools for:

* Reading files (package `readr`)
* Manipulating data frames (packages `dplyr`, `tidyr`, `tibble`)
* Manipulating strings (package `stringr`)
* Manipulating factors (package `forcats`)
* Functional programming (package `purrr`)
* Making visualizations (package `ggplot2`)

The Tidyverse packages are popular but controversial, because some of
them use a syntax different from base R.

RStudio cheat sheets (mostly for Tidyverse packages):

    https://rstudio.com/resources/cheatsheets/



Visualization with `ggplot2`
============================

The fundamental idea of `ggplot2` is that all graphics are composed
of layers.


Documentation:

    https://ggplot2.tidyverse.org/


As an example, let's create a simplifed version of the Best in Show
visualization:

    https://informationisbeautiful.net/visualizations/
        best-in-show-whats-the-top-data-dog/


Layer 1: Data -- the data frame to plot.

Call the `ggplot()` function to set the data layer:
```{r}
library(ggplot2)

ggplot(dogs)
```

Layer 2: GEOMetry -- shapes to represent data.

Add a `geom_` function to set the geometry layer:
```{r}
ggplot(dogs) + geom_point()
```

Layer 3: AESthetic -- "wires" between geometry and data

Call the `aes()` function in `ggplot()` or the `geom_` to set the
mapping between the data layer and geometry layer:
```{r}
ggplot(dogs, aes(x = datadog, y = popularity_all)) + geom_point()
```

Important layers in `ggplot2`:

Layer       | Description
----------  | -----------
data        | A data frame to visualize
aesthetics  | The map or "wires" between data and geometry
geometry    | Geometry to represent the data visually
statistics  | An alternative to geometry
scales      | How numbers in data are converted to numbers on screen
labels      | Titles and axis labels
guides      | Legend settings
annotations | Additional geoms that are not mapped to data
coordinates | Coordinate systems (Cartesian, logarithmic, polar)
facets      | Side-by-side panels


Saving Your Plots
-----------------

Use `ggsave()` to save the last plot you created to a file:

```{r}
ggsave("myplot.png")
```