Where we are and where we’re going

• You might be slightly confused at this point, but…

  • This is new stuff for all of you, and is ‘cutting-edge’ in terms of skills, so take your time, and work hard.
  • We talked about different way of describing numeric variables (mean, variance, standard deviation)
  • You have played around with other types of variables
  • We talked a bit about models, which are imperfect but elegant way of capturing reality
  • We will soon start to write down and estimate some of these models

• Today

  • We will talk about the importance of visual aids in social science
  • We will connect variable types with the way they should be displayed

Why are we discussing this?

``Graphical excellence is that which gives to the viewer the greatest number of ideas in the shortest time with the least ink in the smallest space’’ ― Edward R. Tufte, The Visual Display of Quantitative Information

• Summarize information in a clear and concise way

  • show the distribution of variables

  • show relationships between variables

  • show trends over time

  • show differences between group

  • and much more…

Should we use tables or graphs?

• Data summary is an important part of telling a story in a research paper

  • To do so, we can make a table, or draw a graph
  • Often, it’s better to draw a graph
  • See Kastellec and Leoni (2007) on this point

Table vs. Figure

Let’s look at the movies data

Table

How would you improve that table?

Model table features

• Numbered caption describing table

• Only three horizontal rules

• Formatting for column headers

• No excessive decimal places

Picture - prep

movies_clean<-movies%>%
  mutate(genre = case_when(Action==1 ~ "Action",
                           Animation==1 ~ "Animation",
                           Comedy==1 ~ "Comedy",
                           Drama==1 ~ "Drama",
                           Documentary==1~ "Documentary",
                           Romance==1 ~ "Romance",
                           Short ==1 ~ "Short"))%>%
  drop_na(genre)


movies_plot<-ggplot(data = movies_clean) +
  geom_pointrange(mapping = aes(x = genre, y = rating),
                  stat = "summary",
                  fun.min = function(z) { quantile(z,0.25) },
                  fun.max = function(z) { quantile(z,0.75) },
                  fun = median)+
  coord_flip()+
  theme_minimal()+
  ggtitle("Movie rating by Genre", subtitle = "Lines show interquantile range")

picture

How would you improve that picture?

So…

• Which one is better?

• A table, or a graph?

• I would argue that graphs in general are a better means of communicating your results, summarizing information, and making an argument.

• They require a bit more work, initially.

Let’s talk about figures

Just One variable?

• We must remember our discussion variable types

• Nominal, ordinal (also known as categorical)

• Interval, ratio (also known as continuous)

One categorical variable

• A table, showing frequencies

• Or, a dot plot

Movies example, again:

## # A tibble: 7 x 2
##   genre           n
##   <chr>       <int>
## 1 Action       4688
## 2 Animation    3606
## 3 Comedy      14269
## 4 Documentary  3183
## 5 Drama       16952
## 6 Romance       580
## 7 Short        2724

And plug into table:

A dot plot?

The logic of plotting with ggplot

Lots of helpful material online, see e.g this ggplot intro

• the data being plotted

• a set of mappings from variables in the data to the aesthetics (appearance) of the geometric objects

• the geometric objects (circles, lines, etc.) that appear on the plot

• a coordinate system used to organize the geometric objects

• the facets or groups of data shown in different plots

• possibly, a theme

What you print

movie_freq<-ggplot(data=movies_clean, aes(x=genre))+
  geom_point(stat="count")+
  theme_minimal()+##control how it looks, many themes
  scale_x_discrete("Genre")+
  scale_y_continuous("Frequency")+##label, many other ways to do that
  ggtitle("Movie Frequency by Genre",
          subtitle = "Source: IMDB")+
  coord_flip()##optional

movie_freq # call it

One continuous variable? A histogram?

What you print

movie_hist<-ggplot(movies,aes(x=rating)) + 
  geom_histogram(binwidth=.1)+ ## control the number of movies in each bin. 
  scale_x_continuous("Rating") +
  theme_minimal()+
  ggtitle("Distribution of movies' ratings",
          subtitle = "Source: IMDB")

movie_hist # call it

The histogram

Approximates the distribution of our variable by binning observations into specific values. See below

Two variables

Continuous

What you print

fish_plot<-ggplot(dat,aes(x=GDP90LGN, y = POLITY))+ 
  geom_point() +
  scale_x_continuous("log GDP in 1990 dollars") +
  scale_y_continuous("Polity score")+
  theme_minimal()+
  ggtitle("GDP and Level of Democracy",
          subtitle = "Source: Fish 2012")

fish_plot# call it

One categorical and one continuous variable? A Boxplot?

The boxplot

Captures quartiles and outliers

What you print

opec<- ggplot(data=dat_opec,aes(x=OPEC_member, y = POLITY))+ 
  geom_boxplot()+
  scale_x_discrete("OPEC membership") +
  scale_y_continuous("Polity score") +
  theme_minimal()+
  ggtitle("OPEC membership and Level of Democracy",
          subtitle = "Source: Fish 2012")

opec# call it

Two categorical variables

Maybe a table:

table(dat_mus$OPEC_member, dat_mus$Muslim)##produces contingency tables,such as 2x2

##               
##                Muslim Majority No Muslim majority
##   Member                    10                  1
##   Not a Member              38                108

More than two variables

• Two options:

  • Map one variable into an aesthetic

  • Use small multiples (more than one picture)

Mapping into an aesthetic

What you print

gdp_dem_mus <- ggplot(dat_mus,aes(x=GDP90LGN, y = POLITY,
                                  color = Muslim)) +
  geom_point() +
  scale_x_continuous("log GDP in 1990 dollars") +
  scale_y_continuous("Polity score") +
  theme_classic()+
  ggtitle("GDP and Level of Democracy by Muslim Majority Status",
          subtitle = "Source: Fish 2012")+
  theme_minimal()

Small multiples

What you print

mult<-ggplot(movies_clean,
       aes(x=length, y = rating)) + geom_point() +
  scale_x_continuous("Length") + 
  scale_y_continuous("Rating") +
  facet_wrap(~genre) + 
  theme_minimal()+
  ggtitle("Movies Length and rating by Genre ",
          subtitle = "Source: IMDB")

mult

Principles of Design

Tell the story well

• Give your axes proper labels

• Put the independent variable on the horizontal axis

• Think about what the proper type of plot is

• Don’t use pie charts!

• Don’t use 3D!