Simple (and non-exhaustive) demo of exploratory data analysis tools in R

eda_tools.R

###############################################################################
## brief demo of exploratory data analysis (EDA) tools for data frames in R
## NOTE: the code below is intended to preview the EDA tools ...
## ... it does not exhaustively demonstrate functionality for these tools ...
## ... and it is current as of 2024-04-09 ... 
## ... for more information refer to the documentation for each package
###############################################################################

###############################################################################
## set up
## load tidyverse for data manipulation
## NOTE: tidyverse includes multiple packages ...
## dplyr, tidyr, ggplot2, and more 
library(tidyverse)
## load palmerpenguins for the data itself
library(palmerpenguins)

## what is the :: all about?
penguins
palmerpenguins::penguins
all.equal(penguins, palmerpenguins::penguins)

## what is that strange set of symbols in the code below?
?`%>%`

###############################################################################

## can't we just explore the data "manually"?
## using base R summary to get basic summary stats
penguins %>%
  summary(.)

## using dplyr glimpse() to peek at the first several values of each column
penguins %>%
  glimpse(.)

## counting up missing-ness for a single column with dplyr 
penguins %>%
  summarise(., n_missing_sex = sum(is.na(sex)))

###############################################################################
## purpose-built tools for EDA ...
###############################################################################

###############################################################################
## load the skimr package
library(skimr)

## quickly "skim" the data for an informative summary of all columns
penguins %>%
  skim(.)

###############################################################################

###############################################################################
## load the dlookr package
library(dlookr)

## overall "diagnosis" of variables in the data frame
penguins %>%
  diagnose(.)

## specific summary of categorical variables
penguins %>%
  diagnose_category(.)

## outlier detection and summary
penguins %>%
  diagnose_outlier(.)

## example of using dplyr verbs (e.g., group_by) to refine EDA summaries
penguins %>%
  group_by(., sex) %>%
  diagnose_outlier(., -"sex")

###############################################################################

###############################################################################
## load the naniar package
library(naniar)

## plot of the counts of missing values for each variable
penguins %>%
  gg_miss_var(.)

## example of applying imputation ...
## in this case fixing all imputed values for sex to be "male"
penguins %>%
  mutate(., sex = impute_factor(sex, value = "male"))

## another example of applying imputation ...
## in this case randomly sampling "male" or "female"
penguins %>%
  mutate(., sex = impute_factor(sex, value = sample(c("male","female"), size = 1, prob = c(0.3,0.7))))

###############################################################################

###############################################################################
## next steps
## the tools above are useful for EDA of pre/post-processed data frames
## as an extra step (especially for post-processing) you may need *validation
## a few R packages to consider for data validation ...
## pointblank (https://rstudio.github.io/pointblank/)
## assertr (https://docs.ropensci.org/assertr/)
## validate (https://cran.r-project.org/web/packages/validate/vignettes/cookbook.html)
###############################################################################