Options when building a graph

A wrap up of Visualization

Schwab

Reference Lecture 8 taxonomy of graphics

Graphical parts

  1. Data
  2. Aestheics (variables)
  3. Variable types, and appropriate graphs
  4. Guides
  5. Facets

Data

When building graphs with the tidyverse we want the data to be a dataframe. You can use the class() function to determine if a data file is a dataframe.

library(tidyverse)
#install.packages(mdsr)
library(mdsr)
# In OO class() returns the object type. 
class(iris)
[1] "data.frame"

In this class if a data set is not a data frame (or a tibble) we’ll turn it into one.

Variables

We want to know the variable type (numeric or categorical) we are working with so we can choose the best graph for it.

You can see this by viewing the data.

head(MLB_teams, 1)
# A tibble: 1 × 11
  yearID teamID lgID      W     L  WPct attendance normAttend  payroll metroPop
   <int> <chr>  <fct> <int> <int> <dbl>      <int>      <dbl>    <int>    <dbl>
1   2008 ARI    NL       82    80 0.506    2509924      0.584 66202712  4489109
# ℹ 1 more variable: name <chr>

In general:

  • integers (int) and floats (dbl) are numeric

  • strings (chr) and factors (fct) are categorical

However sometime integers can be categorical, and strings can be numeric. We’ll address these situations as they arrise.

Asthetics

Asthetics are the graph features you want to map to your graph. They are often:

  • x

  • y

  • fill

  • color

  • shape

the pipe |>

The pipe |> is a common r operator that improves code readability. It mean up this into the first argument of that function.

We generally pipe our data into ggplot() when graphing.

Note that metroPop is a numeric variable, and should be graphed with a histogram.

MLB_teams |>   # <- this is the data
    ggplot() +  # <- this is what we are doing
    geom_histogram( mapping = aes( x = metroPop ) )  # <- This is our asthetic

We can change aes()

Using the code above we can easily change aesthetics.

MLB_teams |>   # <- this is the data
    ggplot() +  # <- this is what we are doing
    geom_histogram( mapping = aes( x = payroll ) )  # <- This is our asthetic{r}

Sidenote: scientific notation

We can drop the scientific notation with scale_x_continuous() function.

Code 
MLB_teams |>   # <- this is the data
    ggplot() +  # <- this is what we are doing
    geom_histogram( mapping = aes( x = metroPop ) ) + # <- This is our asthetic
    scale_x_continuous(labels = scales::label_comma())

Faceting

We can also facet our data to make separate plots of features. We have to tell R which variable we want to facet by, I’ll do lgID for league.

Now our x-axis numbers are jumbled. We’ll fix that on the next slide.

Code 
MLB_teams |>   
    ggplot() + 
    geom_histogram( mapping = aes( x = metroPop ) ) + 
    facet_wrap(~lgID) +  # <--  wrap by league
    scale_x_continuous(labels = scales::label_comma())

Sidenote 2: fixing axes numbers

We can abbreviate the axis numbers in a variety of ways. I had to look this one up, because I don’t use it often.

Code 
library(scales) # need this to use cut_short_scale()
MLB_teams |>  
    ggplot() +  
    geom_histogram( mapping = aes( x = metroPop ) ) + 
    facet_wrap(~lgID) +
    scale_x_continuous(
        labels = label_number(scale_cut = cut_short_scale() )
    )

Adjusting alpha

Instead of faceting we can leave the graphs on the same axes but adjust the transparency. alpha is not an aesthetic, but a parameter like mapping. Note: as our lines get longer we break them up.

Code 
library(scales) # need this to use cut_short_scale()
MLB_teams |>  
    ggplot() +  
    geom_histogram( mapping = aes( x = metroPop,
                                 fill = lgID,
                                 ),
                   alpha = 0.5     # <- Alpha is outside aes()
                  ) + 
    
    scale_x_continuous(
        labels = label_number(scale_cut = cut_short_scale() )
    )

Stacked

By default R stacks our plots ( and pick gross colors) This make the American League look taller than the National League. We should dodge them instead by setting the position parameter.

Code 
MLB_teams |>  
    ggplot() +  
    geom_histogram( mapping = aes( x = metroPop,
                                 fill = lgID,
                                 ),
                   alpha = 0.5,
                   position = "dodge"
                  ) + 
    
    scale_x_continuous(
        labels = label_number(scale_cut = cut_short_scale() ),
    )

bin width and bins

Each geom has its own parameters. we can adjust the bin width parameter of geom_histogram() so our bars are not so skinny. We can also specify the exact bins with the bins parameter.

Code 
MLB_teams |>  
    ggplot() +  
    geom_histogram( mapping = aes( x = metroPop,
                                 fill = lgID,
                                 ),
                   alpha = 0.5,
                   position = "dodge",
                   #binwidth = 1000000
                   bins = 10
                   
                  ) + 
    
    scale_x_continuous(
        labels = label_number(scale_cut = cut_short_scale() ),
    )

Changing geom

Once you have your graph, you may want to change the geom to try other visual representations of your data. You can easily do this. Let’s change geom_histogram to geom_density(). We get a warning about bins and position, because those are not parameters of geom_density.

Code 
MLB_teams |>  
    ggplot() +  
    geom_density( mapping = aes( x = metroPop,
                                 fill = lgID,
                                 ),
                   alpha = 0.5,
                   position = "dodge",
                   #binwidth = 1000000
                   bins = 10
                   
                  ) + 
    
    scale_x_continuous(
        labels = label_number(scale_cut = cut_short_scale() ),
    )

geom_density()

geom_density() shows the proportion of your data out of 100.You can change the y-scale to get a better image of what I mean.

Code 
MLB_teams |>  
    ggplot() +  
    geom_density( mapping = aes( x = metroPop,
                                 fill = lgID,
                                 ),
                   alpha = 0.5,
                   position = "dodge",
                   #binwidth = 1000000
                   bins = 10
                   
                  ) + 
    
    scale_x_continuous(
        labels = label_number(scale_cut = cut_short_scale() ),
    )+
    scale_y_continuous(
        labels = label_number(labels = scales::label_comma() )
    )

legend

You can adjust the legend name too.

Code 
MLB_teams |>  
    ggplot() +  
    geom_density( mapping = aes( x = metroPop,
                                 fill = lgID,
                                 ),
                   alpha = 0.5,
                    ) + 
    
    scale_x_continuous(
        labels = label_number(scale_cut = cut_short_scale() ),
    )+
    scale_y_continuous(
        labels = label_number(labels = scales::label_comma() )
    )+
    labs(fill ="League ID")+
    scale_fill_manual(labels = c("American League", "National League"),
                     values = c("blue","red"))

Ethics

Make good graphing choices. Don’t mislead your viewer (especially unintentionally).

Conclusion

There was a lot done during this lecture. I only want you to know what you are able to do. You will not need to do all of this all the time.

You will also be able to use the cheat sheet and AI to help you figure out how to do a certain thing with the tidyverse.