In thisisdaryn/workshop:

options(htmltools.dir.version = FALSE)
knitr::opts_chunk$set(fig.align="center", fig.width=5, fig.height=5, warning = FALSE, message = FALSE)

library(xaringanthemer)
duo_accent(
  primary_color = "ivory",
  secondary_color = "#310A31",
  header_font_google = google_font("Roboto", "400"),
  text_font_google   = google_font("Lato", "300"),
  code_font_family = "Fira Code",
  code_font_url = "https://cdn.rawgit.com/tonsky/FiraCode/1.204/distr/fira_code.css",
  header_color = "#f54278",
  title_slide_text_color = "#354a66"
)

A note

These slides assume your working directory contains the data folder that contains the following files: - Loans.csv - ChicagoEmployees.csv

One way to ensure that you are in that directory is to open the week2.Rproj file. (After downloading the project folder week2 from elearning.)

You can do this in RStudio by File --> Open Project and then navigating to it at the location you downloaded it to.

---

Let's start with the iris data set.

Here are the first few rows of the data frame (using the head function):

head(iris)

Btw, there is also a tail function, that lets you look at the last few rows of the data.

---

Getting a first look at a data set

A function that often comes in handy when first exploring a data set is summary

summary(iris)

-- It provides summary statistics for the columns (variables) in the data frame.

---

Calculating summary statistics on your own:

How would you calculate some of those summary statistics on your own? e.g. for the mean and median of the sepal widths:

--

First, mean:

mean(iris$Sepal.Width)

-- Then, median

median(iris$Sepal.Width)

The mean and the median look very close to each other.

---

Is the distribution symmetric?

We can plot a histogram:

hist(iris$Sepal.Width)

---

We can customise plots using extra parameters:

--

hist(iris$Sepal.Width, xlab = "Sepal Width", main = "Iris Data Set", 
     col = "royalblue") 

---

What about the relationship between Sepal width and length?

To show the relationship between two numerical variables we use plot:

plot(iris$Sepal.Width, iris$Sepal.Length)

This relationship does not appear to be very strong. What is the correlation?

Calculating correlation

We can use cor:

cor(iris$Sepal.Width, iris$Sepal.Length)

-- This is a weakly negative correlation i.e. length seems to generally decrease slightly with increases in width (and vice versa).

---

Is there a relationship between the widths and lengths of the petals?

--

Here's the scatterplot:

plot(iris$Petal.Width, iris$Petal.Length)

--

Ok, this correlation looks pretty strong to me.

cor(iris$Petal.Width, iris$Petal.Length)

---

What about Variance and standard deviation?

Variances and standard deviations of vectors, can be found using the var and sd functions respectively:

var(iris$Petal.Length)

--

sd(iris$Petal.Length)

---

Making boxplots to compare distributions

You can use boxplot to compare numeric variables across the levels of a categorical variable. (The syntax we will use is slightly different to what we have used before):

boxplot(Sepal.Width ~ Species, data = iris)

---

Ok, now to work with data in a file

The list.files function can be used to see the files in a directory.

list.files("data")

--

This shows the 2 data sets in the data directory.

We can read in the loans data set using the function read.csv

#loans <- read.csv("data/Loans.csv", stringsAsFactors = FALSE)

--

The loans data set has 10,000 rows and 17 columns

dim(loans)

If you are using RStudio, you should see it in the Environment window. (And you can browse it in the source pane by clicking on the name loans).

---

Getting counts of categorical variables

An extremely useful function is table which gets counts of categorical variables.

--

Using table to get the counts of different levels of the homeownership variable as in Figure 2.18 of the text.

table(loans$homeownership)

---

Making contingency tables

We can also use table with 2 variables to make a table like the one in Figure 2.17 of the text:

table(loans$app_type, loans$homeownership)

--

The addmargins function can be used to include the totals:

addmargins(table(loans$app_type, loans$homeownership))

---

A Slightly different way to do it

Note that we could have chosen to store the results of the table function to a variable as an intermediate step instead:

ownapps <- table(loans$app_type, loans$homeownership)
addmargins(ownapps)

---

Ok, now to make some bar plots

First get the counts of the homeownership levels:

ownership_counts <- table(loans$homeownership)

--

Now to put them into the barplot function:

barplot(ownership_counts)

---

What about barplots with 2d tables?

--

We can also use the 2d table, ownapps created previously:

--

barplot(ownapps) 

---

Are there other ways to show this bar chart?

--

Real answer: Yes, you can customise plots as much as you want in R

--

For now: we will use the beside argument:

barplot(ownapps, beside = TRUE)

---

A little bit of customising

-- First, we will make a vector of colours. (Most common colour names are available): --

colours <- c("blue", "yellow")

-- Now to apply it to the plot, using the col argument: --

barplot(ownapps, beside = TRUE, col = colours)

---

Adding a legend to the plot

A legend can be added by running the legend command immediately after the barplot function:

barplot(ownapps, beside = TRUE, col = colours)
legend("top", legend = rownames(ownapps), fill = colours)

---

A brief pause: where did rownames come from?

Remember that ownapps was a 2d table.

ownapps

--

You can get the names of the rows in a table using the function rownames:

rownames(ownapps)

--

And you can get the column names using the function colnames:

colnames(ownapps)

---

Using other colours if you wish

Other colours can be selected using html colour codes. For example, we can use a colour scheme close to the text's:

new_colours <- c("#569BBD", "#F4DD00")

Now run similar code as before for the plot:

barplot(ownapps, beside = TRUE, col = new_colours)
legend("top", legend = rownames(ownapps), fill = new_colours)

---

Ending note

• There were quite a few commands used in these slides. Some are more important than others and you will find them recurring more frequently in your use of R.

• for many mathematical functions the corresponding R function is usually easy to figure out e.g. mean, median, sd, min, max

thisisdaryn/workshop documentation built on Jan. 17, 2020, 7:31 p.m.