Research question to plot

Different questions = different visualizations

What is your research question? How can you answer it?

Different kinds of plots let you answer different questions and make different comparisons

One dataset, 25 visualizations:

https://flowingdata.com/2017/01/24/one-dataset-visualized-25-ways/

Same data, (roughly) same question, a bunch of different visualizations:

https://flowingdata.com/2018/10/17/ask-the-question-visualize-the-answer/

Recap of ggplot2

ggplot2

Based on the grammar of graphics

Dataset + variables to coordinate system + geometric object = plot

Geometric objects (“geoms”) are the key piece of the plot

Everything that goes into a plot (including the “ggplot()” part itself) is a function

You can tell by the parentheses at the end
Sometimes you add options (“arguments”) in the parentheses
Sometimes you just have them at the end ()

Basic ggplot template

All screenshots from “Data Visualization” cheatsheet from Rstudio at: https://www.rstudio.com/resources/cheatsheets/

Geom functions

Geoms for 1 variable

Geoms for 2 variables: Continuous X and Y

Geoms for 2 variables: Discrete X, continuous Y

Geoms for 2 variables: Discrete X and Y

Geoms for 2 variables: Continuous bivariate distribution

Geoms for 2 variables: Continuous function

Geoms for 2 variables: Error / variability

Four different ways to include error bars

You must specify the min and max for the error bars

Geoms for 3 variables

Geom primitives

Line segments

Stat functions

Statistical functions

Each geom has a default statistic associated with it

e.g., geom_bar uses the “count” statistic

But you can explicitly change the statistic

e.g., instead of the count (frequency) use a proportion that you’ve calculated and use the “identity” statistic with the proportion variable

Position functions

How to arrange geoms that would otherwise overlap

Arguments in the geom_bar (and similar) geoms:

geom_bar(position = “dodge”)

Move bars to be next to one another

geom_bar(position = “fill”)

Fill up all vertical space with the bar, split appropriately between categories

geom_bar(position = “stack”)

Stack the bars on top of one another, but don’t fill up the whole vertical space

Position functions

This can replace geom_point

geom_jitter()

Add random noise on both x and y axes to avoid overplotting
Can be modified to only jitter on 1 or the other dimension:
- geom_jitter(width = 0.5)
- geom_jitter(height = 0.5)
Or to limit how far the jitter spreads:
- geom_jitter(width = 0.5, height = 0.5)

Coordinate functions

Coordinate system for the plot

coord_cartesian()

default for most plots

coord_fixed(ratio = x/y)

change aspect ratio between x and y (argument)

coord_flip()

Make x into y and y into x

coord_polar()

Polar coordinates

coord_trans(ytrans = “sqrt”)

Transform one of the axes (argument), such as the y axis being square rooted

Facet functions

Faceting functions

There are a few faceting functions, but one does most of the things you will ever need: facet_grid()

facet_grid(cols = vars(X1))

Can facet into columns based on a variable (here, X1)

facet_grid(rows = vars(X2))

Can facet into rows based on a variable (here, X2)

facet_grid(rows = vars(X2), cols = vars(X1))

Can facet into both rows and columns based on two variables

Theme functions

Themes are sets of changes to the overall look of the plot

You just apply the theme and all the changes are made

You can use some built-in default themes (today), or use packaged themes that other people have made (soon) or even make your own theme (just for fun)