w3_data_visualization.Rmd

---
title: "Week 3 - Data Visualization"
author: "Nicholas Good"
output:
  rmarkdown::html_document:
    toc: true
    toc_float: true
    theme: yeti
---

```{r global_options, include=FALSE}
  knitr::opts_chunk$set(echo=TRUE, warning=FALSE, message=FALSE, results = 'hide')
```

---

# Material

This week's slides and data are availble on the [shared google drive folder](https://tinyurl.com/R-for-fire-data).

# Packages

This exercise uses the package `ggplot2`. The `ggplot2` package is part of the `tidyverse` collection of packages.

```{r packages}
 # install.packages("ggplot2")
 # install.packages("tidyverse")
 # uncomment the code above to install:
 # ggplot2 and/or all the tidyverse packages
 # library(ggplot2) # to include the ggplot library
 # install.packages("magrittr")
 # comment the code above to install: magrittr
  
  library(tidyverse) # include tidyverse packages
  library(magrittr)
```

---

# Exploring data graphically

Plotting data is an important tool for exploratory data analysis. Many object types support a `plot()` method for quick visualization of results. The `plot()` function can also be used for custom plots, however it is fairly limited in terms of its scope. The `ggplot2` package provides a much more comprehensive and customizable set of tools for data visualization.

---

# ggplot2

The [ggplot2](https://www.rstudio.com/wp-content/uploads/2015/03/ggplot2-cheatsheet.pdf) package is built on an underlying [grammer of graphics](http://vita.had.co.nz/papers/layered-grammar.pdf), that describes how data is transformed into a visual statistical representation. Some of the core elements are:

 * **data** - information that you will **map** to aesthetic attributes
 
 * **layers** - the geometric (**geoms**) elements that make you use to visualize statistical data transformations
 
 * **scales/coordinates** - determine how data is mapped
 
 * **facets** - a process of splitting and visualizing subsets of data
 
 * **themes** - methods to control details of a graphics appearance

# This week's data

Let's load the [arctas](https://drive.google.com/open?id=0B5Ag5T2bIuQzNUFJNHN3V0dYVjg) data set from the first week. Remember to tweak the *path* if needed.

```{r load_data}
 # create object containing a vector of file names:
  files <- list.files("../data/arctas",
                      full.names = TRUE,
                      pattern = "R14.ict$")

 # load files in the file list
  data_arctas <- lapply(files,
                 read_csv, skip = 387,
                          col_types = cols()) %>%
                 bind_rows()

                 names(data_arctas) %<>% tolower

 # clean the data
  data_arctas[data_arctas == -999999999 | data_arctas == -888888888] <- NA

 # grab the first flight
  data_first_flight <- filter(data_arctas,
                              flight == first(unique(flight)))
```

Check the data looks reasonable using the functions below. Is there anything else you would check?

```{r check_data}
  head(data_arctas, 2)

  nrow(data_arctas)

  colnames(data_arctas)

  summary(data_arctas)

  class(data_arctas)
```

---

# Basics

As a minimum, each `ggplot` graphic requires at least three elements:

* data
* aesthetic
* layer (geom)

```{r basics}
  ggplot(data = data_first_flight,
                mapping = aes(x = utc, y = o3)) +
  geom_point()
```

We can add additional *geoms* to a `ggplot` object, and modify them:

```{r add_geoms}
  ggplot(data = data_first_flight,
         mapping = aes(x = utc, y = o3)) +
  geom_point() +
  geom_point(aes(y = so2), color = "green", shape = 1)
```

* note that unmodified arguments are inherited from the original `ggplot(...)` function.

* are there any problems with this plot?

# Formating data for ggplot

The `ggplot` package often prefers data in a longer format. The `tidyr` package contains useful tools for converting to longer and wider formats:

```{r, data_width}
  library(dplyr)
  library(tidyr)

  long_data <- dplyr::select(data_first_flight, utc, o3, so2) %>%
               tidyr::gather("var", "val", 2:3)

  ggplot(long_data, aes(utc, val, color = var)) +
  geom_point()
```

---

# Plots as objects

Plots in R are objects, therefore you can store can assign a plot to an object name for manipulation at a later time:

```{r, plot_objects}
  p <- ggplot(data = data_first_flight,
              mapping = aes(x = utc, y = o3)) +
       geom_point()
  p
```

---

# Basic plots

The `ggplot` package has numerous *geoms* for creating common plot types.

* Use `geom_hist()` to create a histogram of the *temperature* variable for flight #10.

* Use `geom_box()` to create a box plot of the *noy* variable with a box for each flight. You'll need to convert the x variable *flight* to a categorical *class* e.g. using the `as.factor()` function.

---

# Faceting

You will often want to plot groups of data separately. The `facet_grid()` and `facet_wrap()` *geoms* enable you to do this:

```{r facets, fig.width=12, fig.height=30}
  p <- ggplot(data = data_arctas, mapping = aes(x = utc, y = o3)) +
       geom_point() +
       facet_grid(flight ~ .)
```

* Try the faceting by some different variables.

* Apply the `scales = "free"` argument to `facet_grid()`.

---

# Ordering

Ordering data in a plot can be very informative:

```{r ordering}
  p_data <- select(data_arctas, flight, no) %>%
            group_by(flight) %>%
            summarise(mean_no = mean(no, na.rm = TRUE))

  p <- ggplot(p_data, aes(flight, mean_no)) +
       geom_point()
```

```{r plot_ordered}
  p_data <- select(data_arctas, flight, no) %>%
            group_by(flight) %>%
            summarise(mean_no = mean(no, na.rm = TRUE)) %>%
            arrange(mean_no) %>%
            mutate(rank = row_number())

  p <- ggplot(p_data, aes(rank, mean_no)) +
       geom_point()
```

---

# Lines

You can fit curves to your data using `ggplot()` and `geom_smooth()`:

```{r add_lines}
  p <- ggplot(p_data, aes(rank, mean_no)) +
       geom_point() +
       geom_smooth()
```

* try passing some different arguments to `geom_smooth()`

---

# Text

You can annotate you plots using `geom_text()`. Each *geom* can take a number of additional arguments to alters its appearance.

```{r add_text}
  p <- ggplot(p_data, 
              aes(rank, mean_no)) +
       geom_point() +
       geom_smooth() +
       geom_text(aes(label = flight),
                 hjust = 0.5, vjust = -0.5)
```

---

# Appearance

* detailed plots appearance is controlled by the `theme()` function. Try changing the appearance of a plot using the `theme()` function, for example the axis font size and orientation. You can use the help or search online for examples.

* you can apply general themes, which are included in base R or for example from the `ggthemes()` package. Install the `ggthemes()` package and try applying some different themes. Start by typing `theme_` into the console and scroll through the options.

* the `ggplot()` package has a number of built in functions for modifying how graphics are displayed try some of the these functions (these often replicate arguments you can pass to the `theme()` function.

```{r helper_functions, eval = FALSE}
 xlab()
 ylab()
 scale_y_log10()
 ggtitle()
```

---

# Interactive graphics

Often graphics were you can read off the numbers are useful. The [plotly](https://plot.ly/r/) packages allows conversion of `ggplot` objects to interactive `plotly` objects

```{r plotly}
 # install.packages("plotly")
 # include the plotly library
  library(plotly)

 # create a ggplot object
  p <- ggplot(p_data,
              aes(rank, mean_no)) +
       geom_point() +
       geom_smooth()

 # convert to a plotly object
  p_plotly  <- ggplotly(p)

 # display the plot
  p_plotly
```

---