In luisDVA/unheadr: Handle Data with Messy Header Rows and Broken Values

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

unheadr

The goal of unheadr is to help wrangle data when it has embedded subheaders, or when values are wrapped across several rows. https://unheadr.liomys.mx/

Installation

You can install the CRAN release or the development version with:

# Install unheadr from CRAN:
install.packages("unheadr")

# Or install the development version from GitHub with:
# install.packages("remotes")
remotes::install_github("luisDVA/unheadr")

The reasoning behind the package and some of the possible uses of unheadr are described in this publication:

Verde Arregoitia, L. D., Cooper, N., D'Elía, G. (2018). Good practices for sharing analysis-ready data in mammalogy and biodiversity research. Hystrix, the Italian Journal of Mammalogy, 29(2), 155-161. Open Access, DOI 10.4404/hystrix-00133-2018

Usage

Load the package first.

library(unheadr)

Main functions

untangle2()

untangle2() puts embedded subheaders into their own variable, using regular expressions to identify them.

In the data below (a subset of a bundled dataset which can be loaded with data(primates2017)), there are rows that correspond to values in grouping variables. These should be in their own column. Instead, they are embedded within the data rectangle. This is a common practice in many disciplines. This data presentation looks OK and is easy to read, but hard to work with (for example: calculating group-wise summaries).

In this example, values for an implicit "geographic region" variable and an implicit "taxonomic family" variable are embedded in the column that contains the observational units (the scientific names of various primates).

|scientific_name |common_name |red_list_status | mass_kg| |:----------------------------|:----------------------------|:---------------|-------:| |Asia |NA |NA | NA| |CERCOPITHECIDAE |NA |NA | NA| |Trachypithecus obscurus |Dusky Langur |NT | 7.13| |Presbytis sumatra |Black Sumatran Langur |EN | 6.00| |Rhinopithecus roxellana |Golden Snub-nosed Monkey |EN | NA| |HYLOBATIDAE |NA |NA | NA| |Hylobates funereus |East Bornean Gray Gibbon |EN | NA| |Hylobates klossii |Kloss's Gibbon |EN | 5.80| |Nomascus concolor |Western Black Crested Gibbon |CR | 7.71|

For a tidier structure, the subheaders embedded in the scientific_name column need to be plucked out and placed in their own variable. This was initially the main objective of unheadr and what untangle2() was made for. The function can be used with magrittr pipes as a dplyr-type verb.

If these subheaders can be matched in bulk with a regular expression because they share a prefix, suffix, or anything in common, we can save a lot of time. Otherwise, they can be matched by name. For more details, see the examples and vignette.

The 'untangled' version of the data:

|scientific_name |common_name |red_list_status | mass_kg|family |region | |:----------------------------|:----------------------------|:---------------|-------:|:---------------|:----------| |Trachypithecus obscurus |Dusky Langur |NT | 7.13|CERCOPITHECIDAE |Asia | |Presbytis sumatra |Black Sumatran Langur |EN | 6.00|CERCOPITHECIDAE |Asia | |Rhinopithecus roxellana |Golden Snub-nosed Monkey |EN | NA|CERCOPITHECIDAE |Asia | |Hylobates funereus |East Bornean Gray Gibbon |EN | NA|HYLOBATIDAE |Asia | |Hylobates klossii |Kloss's Gibbon |EN | 5.80|HYLOBATIDAE |Asia | |Nomascus concolor |Western Black Crested Gibbon |CR | 7.71|HYLOBATIDAE |Asia |

Now we can easily perform grouping operations and summarize the data (e.g. calculating average body mass by Family).

unbreak_vals()

This function uses regex to fix values that are broken across two rows. This usually happens when we are formatting a table and we need to fit it on a page.

# Set up a toy dataset
dogsDesc <-
  data.frame(
    stringsAsFactors = FALSE,
    dogs = c(
      "Retriever", "(Golden)",
      "Retriever", "(Labrador)", "Bulldog", "(French)"
    ),
    coat = c("long", NA, "short", NA, "short", NA)
  )

dogsDesc

We can match the opening brackets with regex.

unbreak_vals(df = dogsDesc, regex = "^\\(", ogcol = dogs, newcol = dogs_unbroken)

unwrap_cols()

Use this function to unwrap and glue values that have been wrapped across multiple rows for presentation purposes, with an inconsistent number of empty or NA values padding out the columns.

# Set up the data
nyk <-
  data.frame(
    stringsAsFactors = FALSE,
    player = c(
      "Marcus Camby", NA, NA,
      NA, NA, NA, NA, "Allan Houston", NA,
      "Latrell Sprewell", NA, NA
    ),
    listed_height_m. = c(
      2.11, NA, NA, NA, NA, NA,
      NA, 1.98, NA, 1.96, NA, NA
    ),
    teams_chronological = c(
      "Raptors", "Knicks",
      "Nuggets", "Clippers", "Trail Blazers",
      "Rockets", "Knicks", "Pistons",
      "Knicks", "Warriors", "Knicks",
      "Timberwolves"
    ),
    position = c(
      "Power forward", "Center",
      NA, NA, NA, NA, NA,
      "Shooting guard", NA, "Small forward", NA, NA
    )
  )
nyk

Unwrap the elements in the variable that defines the groups, separating with commas.

unwrap_cols(nyk, groupingVar = player, separator = ", ")

unbreak_rows()

This function merges sets of two contiguous rows upwards by pasting the values of the lagging row to the values of the leading row (identified using regular expressions).

The following table of basketball records has two sets of header rows with values broken across two contiguous rows.

bball <- data.frame(
  stringsAsFactors = FALSE,
  v1 = c(
    "Player", NA, "Sleve McDichael", "Dean Wesrey",
    "Karl Dandleton", "Player",
    NA,
    "Mike Sernandez",
    "Glenallen Mixon",
    "Rey McSriff"
  ),
  v2 = c(
    "Most points", "in a game", "55", "43", "41", "Most varsity",
    "games played", "111", "109",
    "104"
  ),
  v3 = c(
    "Season", "(year ending)", "2001", "2000", "2010", "Season",
    "(year ending)", "2005",
    "2004", "2002"
  )
)

unbreak_rows() merges these rows if we can match them with a common pattern.

# Match with regex on variable v2
unbreak_rows(bball, regex = "^Most", ogcol = v2)

mash_colnames()

When column names are broken up across the top n rows of a data frame or tibble, mash_colnames() makes many header rows into column names. Existing names can be kept or ignored.

# Data with broken headers
babies <-
  data.frame(
    stringsAsFactors = FALSE,
    Baby = c(NA, NA, "Angie", "Yean", "Pierre"),
    Age = c("in", "months", "11", "9", "7"),
    Weight = c("kg", NA, "2", "3", "4"),
    Ward = c(NA, NA, "A", "B", "C")
  )

babies

# Mash, including the object names
mash_colnames(babies, n_name_rows = 2, keep_names = TRUE)

For inputs with ragged column names (NA values in the first row), the first row can be filled row-wise before mashing.

# Data with ragged headers
survey <-
  data.frame(
    stringsAsFactors = FALSE,
    X1 = c("Participant", NA, "12", "34", "45", "123"),
    X2 = c(
      "How did you hear about us?",
      "TV", "TRUE", "FALSE", "FALSE", "FALSE"
    ),
    X3 = c(NA, "Social Media", "FALSE", "TRUE", "FALSE", "FALSE"),
    X4 = c(NA, "Radio", "FALSE", "TRUE", "FALSE", "TRUE"),
    X5 = c(NA, "Flyer", "FALSE", "FALSE", "FALSE", "FALSE"),
    X6 = c("Age", NA, "31", "23", "19", "24")
  )
survey

# Ignoring names and using sliding headers
mash_colnames(survey,2,keep_names = FALSE,sliding_headers = TRUE, sep = "_")

annotate_mf() and annotate_mf_all()

Sometimes embedded subheaders can't be matched by content or context, but they share the same formatting in a spreadsheet file.

annotate_mf() flattens four common approaches to confer meaningful formatting to cells and adds this as a character string to the target variable.

example_spreadsheet <- system.file("extdata/dog_test.xlsx", package = "unheadr")
annotate_mf(example_spreadsheet,orig = Task, new=Task_annotated)

annotate_mf_all() applies the same approach to all values in the dataset.

example_spreadsheet_all <- system.file("extdata/boutiques.xlsx", package = "unheadr")
annotate_mf(example_spreadsheet_all)

Lastly, regex_valign() can adjust the whitespace (padding) within a character vector with one element per line, for easier parsing with readr.

guests <- 
  unlist(strsplit(c("6     COAHUILA        20/03/2020
712        COAHUILA             20/03/2020"),"\n"))
guests
regex_valign(guests, "\\b(?=[A-Z])")

The inconsistent whitespace between the elements in each line can be adjusted after matching a position of interest through regular expressions.

luisDVA/unheadr documentation built on Aug. 16, 2022, 5:28 a.m.