skim {skimr}

Skim a data frame, getting useful summary statistics

Description

skim() is an alternative to summary(), quickly providing a broad overview of a data frame. It handles data of all types, dispatching a different set of summary functions based on the types of columns in the data frame.

Usage

skim(data, ..., .data_name = NULL)

skim_tee(data, ..., skim_fun = skim)

skim_without_charts(data, ..., .data_name = NULL)

Arguments

data

A tibble, or an object that can be coerced into a tibble.

...

Columns to select for skimming. When none are provided, the default is to skim all columns.

.data_name

The name to use for the data. Defaults to the same as data.

skim_fun

The skim function used.

skim

The skimming function to use in skim_tee().

Details

Each call produces a skim_df, which is a fundamentally a tibble with a special print method. One unusual feature of this data frame is pseudo- namespace for columns. skim() computes statistics by data type, and it stores them in the data frame as ⁠<type>.<statistic>⁠. These types are stripped when printing the results. The "base" skimmers (n_missing and complete_rate) are the only columns that don't follow this behavior. See skim_with() for more details on customizing skim() and get_default_skimmers() for a list of default functions.

If you just want to see the printed output, call skim_tee() instead. This function returns the original data. skim_tee() uses the default skim(), but you can replace it with the skim argument.

The data frame produced by skim is wide and sparse. To avoid type coercion skimr uses a type namespace for all summary statistics. Columns for numeric summary statistics all begin numeric; for factor summary statistics begin factor; and so on.

See partition() and yank() for methods for transforming this wide data frame. The first function splits it into a list, with each entry corresponding to a data type. The latter pulls a single subtable for a particular type from the skim_df.

skim() is designed to operate in pipes and to generally play nicely with other tidyverse functions. This means that you can use tidyselect helpers within skim to select or drop specific columns for summary. You can also further work with a skim_df using dplyr functions in a pipeline.

Value

A skim_df object, which also inherits the class(es) of the input data. In many ways, the object behaves like a tibble::tibble().

Customizing skim

skim() is an intentionally simple function, with minimal arguments like summary(). Nonetheless, this package provides two broad approaches to how you can customize skim()'s behavior. You can customize the functions that are called to produce summary statistics with skim_with().

Unicode rendering

If the rendered examples show unencoded values such as ⁠<U+2587>⁠ you will need to change your locale to allow proper rendering. Please review the Using Skimr vignette for more information (vignette("Using_skimr", package = "skimr")).

Otherwise, we export skim_without_charts() to produce summaries without the spark graphs. These are the source of the unicode dependency.

Examples

skim(iris)

# Use tidyselect
skim(iris, Species)
skim(iris, starts_with("Sepal"))
skim(iris, where(is.numeric))

# Skim also works groupwise
iris %>%
  dplyr::group_by(Species) %>%
  skim()

# Which five numeric columns have the greatest mean value?
# Look in the `numeric.mean` column.
iris %>%
  skim() %>%
  dplyr::select(numeric.mean) %>%
  dplyr::top_n(5)

# Which of my columns have missing values? Use the base skimmer n_missing.
iris %>%
  skim() %>%
  dplyr::filter(n_missing > 0)

# Use skim_tee to view the skim results and
# continue using the original data.
chickwts %>%
  skim_tee() %>%
  dplyr::filter(feed == "sunflower")

# Produce a summary without spark graphs
iris %>%
  skim_without_charts()