Load some data

library(tidyverse)

## Warning: package 'readr' was built under R version 4.1.1

library(ggformula)
library(palmerpenguins)

theme_set(theme_bw(base_size = 16))

head(penguins)

## # A tibble: 6 × 8
##   species island bill_length_mm bill_depth_mm flipper_length_… body_mass_g sex  
##   <fct>   <fct>           <dbl>         <dbl>            <int>       <int> <fct>
## 1 Adelie  Torge…           39.1          18.7              181        3750 male 
## 2 Adelie  Torge…           39.5          17.4              186        3800 fema…
## 3 Adelie  Torge…           40.3          18                195        3250 fema…
## 4 Adelie  Torge…           NA            NA                 NA          NA <NA> 
## 5 Adelie  Torge…           36.7          19.3              193        3450 fema…
## 6 Adelie  Torge…           39.3          20.6              190        3650 male 
## # … with 1 more variable: year <int>

Here is a histogram exploring the body mass of the penguins.

gf_histogram(~ body_mass_g, data = penguins, color = 'black') %>%
  gf_labs(x = "Body Mass (in grams)")

## Warning: Removed 2 rows containing non-finite values (stat_bin).

What are the weaknesses of the histogram?

gf_histogram(~ body_mass_g, data = penguins, color = 'black', 
             binwidth = 50) %>%
  gf_labs(x = "Body Mass (in grams)")

## Warning: Removed 2 rows containing non-finite values (stat_bin).

An alternative is the density curve which uses kernel density estimation to get a curve. The details of the kernel density estimation isn’t overly important, and it is possible to change the estimation. The default estimation works pretty well however.

gf_density(~ body_mass_g, data = penguins) %>%
  gf_labs(x = "Body Mass (in grams)")

## Warning: Removed 2 rows containing non-finite values (stat_density).

Multivariate Thinking

In general, exploring univariate distributions are important, but in most situations, it is also important to explore these distributions in a multivariate framework. This means, exploring the distribution of the outcome attribute by other attributes.

** Insert interactive components**

The boxplot is one way to do this.

gf_boxplot(@@ ~ body_mass_g, data = penguins) %>%
  gf_labs(x = "Body Mass (in grams)",
          y = "$$")

Boxplots are simple representations, but since they are only based on 5 numbers, can be too simple.

Density plots for each group could be explored, but can get difficult to interpret with many groups. Violin plots (or a related sina plot) are the solution.

gf_violin(@@ ~ body_mass_g, data = penguins, fill = 'gray80') %>%
  gf_labs(x = "Body Mass (in grams)",
          y = "$$")

gf_violin(@@ ~ body_mass_g, data = penguins, fill = 'gray80',
          draw_quantiles = c(0.1, 0.5, 0.9)) %>%
  gf_labs(x = "Body Mass (in grams)",
          y = "$$")

library(ggforce)

gf_sina(body_mass_g ~ @@, data = penguins) %>%
  gf_labs(y = "Body Mass (in grams)",
          x = "$$") %>%
  gf_refine(coord_flip())

gf_violin(body_mass_g ~ species, data = penguins) %>%
  gf_sina(body_mass_g ~ species, data = penguins) %>%
  gf_labs(y = "Body Mass (in grams)",
          x = "$$") %>%
  gf_refine(coord_flip())