datawizard

datawizard is a lightweight R package designed for data manipulation and statistical transformations. It is part of the easystats ecosystem.

It focuses on two main areas:

1. Data manipulation: Provides functions similar to tidyverse (dplyr, tidyr) for selecting, filtering, and reshaping data. Key characteristics include:
   • All functions use the data_* prefix for easy identification.
   • Functions are "string-friendly," making them easy to use within other functions or programmatic workflows.
   • Extremely lightweight with no dependencies.
2. Statistical transformations: Includes powerful functions for common transformations such as standardization, normalization, rescaling, rank-transformation, scale reversing, recoding, and binning.

The {datawizard} package is designed for seamless integration into R pipe-workflows (|>). Most functions follow a consistent argument pattern:

1. The first argument is always the data.
2. For functions operating on data frames, the next two arguments are select and exclude for variable manipulation.
3. Subsequent arguments are task-specific parameters.

Functions that accept data frames typically return a modified data frame, allowing you to chain multiple operations together.

iris |>
  data_filter(Species %in% c("versicolor", "virginica")) |>
  data_select(contains("\\.")) |>
  data_relocate(ends_with("Length")) |>
  data_remove(4) |>
  head()

After data preparation, use {datawizard} to apply preprocessing steps and statistical transformations required by specific statistical tests. These functions allow you to modify variable distributions to meet model assumptions.

Key functions include:

• standardize(): Center and scale data.
• normalize(): Scale variables to a 0-1 range.
• adjust(): Adjust data for the effect of other variables.
• slide(): Shift numeric value ranges.
• ranktransform(): Convert numeric values to integer ranks.

d <- data.frame(
  a = c(-2, -1, 0, 1, 2),
  b = c(3, 4, 5, 6, 7)
)

# Standardize data with specific center and scale values
standardize(d, center = c(3, 4), scale = c(2, 4))

Use {datawizard} to clean and reshape raw data for statistical modeling. The package provides functions for selecting observations, selecting variables, renaming, and reshaping data. The syntax is designed to be similar to the {tidyverse} ecosystem.

Key functions include:

• data_filter(): Select specific observations.
• data_select(): Select specific variables.
• data_extract(): Extract a single variable.
• data_rename(): Rename variables.
• data_to_long(): Convert data from wide to long format.
• data_to_wide(): Convert data from long to wide format.
• data_join(): Join two data frames.

stocks <- data.frame(
  time = as.Date("2009-01-01") + 0:4,
  X = rnorm(5, 0, 1),
  Y = rnorm(5, 0, 2)
)

# Convert wide data to long format
data_to_long(
  stocks,
  select = -c("time"),
  names_to = "stock",
  values_to = "price"
)

You can install datawizard from CRAN, the easystats r-universe, or directly from GitHub.

Tip: Instead of loading library(datawizard), it is recommended to use library(easystats) to make all features of the easystats ecosystem available. To keep your packages updated, use easystats::install_latest().

# Release version from CRAN
install.packages("datawizard")

# Development version from r-universe
install.packages("datawizard", repos = "https://easystats.r-universe.dev")

# Development version from GitHub
remotes::install_github("easystats/datawizard")

The {datawizard} functions follow a consistent design principle to facilitate easy learning and seamless integration into R pipe-workflows (|>).

1. Argument Order: The first argument is always the data.
2. Data Frame Methods: For functions working on data frames, the next two arguments are typically used to select and exclude variables.
3. Task Arguments: Subsequent arguments are specific to the function's task.
4. Pipe-friendly: Functions that accept data frames usually return a modified data frame, allowing them to be chained together.

iris |>
  data_filter(Species %in% c("versicolor", "virginica")) |
  data_select(contains("\.")) |
  data_relocate(ends_with("Length")) |
  data_remove(4) |
  head()

You can combine various {datawizard} functions to transform your data in a single workflow. Common operations include:

• data_filter(): Filter rows based on conditions.
• data_select(): Select specific columns (often using pattern matching like contains()).
• data_relocate(): Change the order of columns.
• data_remove(): Remove columns by index or name.

Example workflow:

iris |>
  data_filter(Species %in% c("versicolor", "virginica")) |>
  data_select(contains("\\.")) |>
  data_relocate(ends_with("Length")) |>
  data_remove(4) |>
  head()

Use describe_distribution() to generate a comprehensive descriptive summary for all variables in a dataframe or for a single specific variable. The output includes statistics such as Mean, SD, IQR, Range, Skewness, Kurtosis, sample size (n), and missing values (n_Missing).

# For a whole dataframe
data(iris)
describe_distribution(iris)

# For a single variable
describe_distribution(mtcars$wt)

Use data_relocate() to change the position of columns and data_rename() to change their names.

• data_relocate(df, select, before): Moves the specified select columns to a position before the before column.
• data_rename(df, old_names, new_names): Renames columns by passing a vector of current names and a vector of new names.

# Move 'Species' to the front, before 'Sepal.Length'
data_relocate(iris, select = "Species", before = "Sepal.Length")

# Rename specific columns
data_rename(iris, c("Sepal.Length", "Sepal.Width"), c("length", "width"))

The package includes several functions for common statistical data transformations:

• standardize(df): Performs z-score standardization on numeric data.
• winsorize(df): Applies winsorization to handle outliers.
• center(df): Performs grand-mean centering.
• ranktransform(df): Applies a rank transformation to the data.
• change_scale(vec, new_min, new_max): Rescales a numeric vector to a new specified range (defaults to 0-100 if not specified).
• categorize(vec, split, n_groups): Recodes data into groups. For example, using split = "quantile" to create groups based on quantiles.

# Standardize data
standardize(swiss)

# Winsorize data
winsorize(anscombe)

# Center data
center(anscombe)

# Rescale a vector
change_scale(c(0, 1, 5, -5, -2))

# Categorize into quantiles
categorize(x, split = "quantile", n_groups = 3)

The describe_distribution() function provides a comprehensive summary of data properties and distributions. It computes various indices for each variable in a data frame, including measures of centrality, dispersion, range, skewness, and kurtosis.

Output includes:

• Mean, SD, IQR
• Min, Max
• Skewness, Kurtosis
• Sample size (n) and missing values (n_Missing)

describe_distribution(mtcars)

The smoothness() function computes additional data properties. You can specify the calculation method using the method argument (e.g., method = "diff").

x <- (-10:10)^3 + rnorm(21, 0, 100)
smoothness(x, method = "diff")