Data Structures, Summaries, and Visualisations for Missing Data

Missing values are ubiquitous in data and need to be explored and handled in the initial stages of analysis. 'naniar' provides data structures and functions that facilitate the plotting of missing values and examination of imputations. This allows missing data dependencies to be explored with minimal deviation from the common work patterns of 'ggplot2' and tidy data.


AppVeyor Build Status Travis-CI Build Status Coverage Status

naniar provides principled, tidy ways to summarise, visualise, and manipulate missing data with minimal deviations from the workflows in ggplot2 and tidy data. It does this by providing:

  • Shadow matrices, a tidy data structure for missing data (as_shadow()).
  • Shorthand summaries for missing data: n_miss()/n_complete(); prop_miss()/prop_complete().
  • Numerical summaries of missing data in variables (miss_var_summary(), miss_var_run), and cases (miss_case_summary(), miss_case_table()).
  • Visualisation methods: e.g., geom_miss_point(), gg_miss_var(),

For more details on the workflow and theory underpinning naniar, read the vignette Getting started with naniar.

For a short primer on the data visualisation available in naniar, read the vignette Gallery of Missing Data Visualisations.

Installation

Currently naniar is only available on github

devtools::install_github("njtierney/naniar")

A short overview of naniar

Visualising missing data might sound a little strange - how do you visualise something that is not there? One approach to visualising missing data comes from ggobi and manet, where we replace "NA" values with values 10% lower than the minimum value in that variable. This is provided with the geom_miss_point() ggplot2 geom, which we can illustrate by exploring the relationship between Ozone and Solar radiation from the airquality dataset.

 
library(ggplot2)
 
ggplot(data = airquality,
       aes(x = Ozone,
           y = Solar.R)) +
  geom_point()
#> Warning: Removed 42 rows containing missing values (geom_point).

ggplot2 does not handle these missing values, and we get a warning message about the missing values.

We can instead use the geom_miss_point() to display the missing data

 
library(naniar)
 
ggplot(data = airquality,
       aes(x = Ozone,
           y = Solar.R)) +
  geom_miss_point()

geom_miss_point() has shifted the missing values to now be 10% below the minimum value. The missing values are a different colour so that missingness becomes pre-attentive. As it is a ggplot2 geom, it supports features like faceting and other ggplot features.

 
p1 <-
ggplot(data = airquality,
       aes(x = Ozone,
           y = Solar.R)) + 
  geom_miss_point() + 
  facet_wrap(~Month, ncol = 2) + 
  theme(legend.position = "bottom")
 
p1

Data Structures

naniar provides a data structure for working with missing data, the shadow matrix (Swayne and Buja, 1998). The shadow matrix is the same dimension as the data, and consists of binary indicators of missingness of data values, where missing is represented as “NA”, and not missing is represented as “!NA”, and variable names are kep the same, with the added suffix “_NA" to the variables.

 
head(airquality)
#>   Ozone Solar.R Wind Temp Month Day
#> 1    41     190  7.4   67     5   1
#> 2    36     118  8.0   72     5   2
#> 3    12     149 12.6   74     5   3
#> 4    18     313 11.5   62     5   4
#> 5    NA      NA 14.3   56     5   5
#> 6    28      NA 14.9   66     5   6
 
as_shadow(airquality)
#> # A tibble: 153 x 6
#>    Ozone_NA Solar.R_NA Wind_NA Temp_NA Month_NA Day_NA
#>      <fctr>     <fctr>  <fctr>  <fctr>   <fctr> <fctr>
#>  1      !NA        !NA     !NA     !NA      !NA    !NA
#>  2      !NA        !NA     !NA     !NA      !NA    !NA
#>  3      !NA        !NA     !NA     !NA      !NA    !NA
#>  4      !NA        !NA     !NA     !NA      !NA    !NA
#>  5       NA         NA     !NA     !NA      !NA    !NA
#>  6      !NA         NA     !NA     !NA      !NA    !NA
#>  7      !NA        !NA     !NA     !NA      !NA    !NA
#>  8      !NA        !NA     !NA     !NA      !NA    !NA
#>  9      !NA        !NA     !NA     !NA      !NA    !NA
#> 10       NA        !NA     !NA     !NA      !NA    !NA
#> # ... with 143 more rows

Using the shadow matrix helps you manage where missing values are in your dataset and make it easy to do visualisations where you split by missingness:

 
airquality %>%
  bind_shadow() %>%
  ggplot(aes(x = Temp,
             fill = Ozone_NA)) + 
  geom_density()

And even visualise imputations

 
airquality %>%
  bind_shadow() %>%
  simputation::impute_lm(Ozone ~ Temp + Solar.R) %>%
  ggplot(aes(x = Solar.R,
             y = Ozone,
             colour = Ozone_NA)) + 
  geom_point()
#> Warning: Removed 7 rows containing missing values (geom_point).

naniar does this while following consistent principles that are easy to read, thanks to the tools of the tidyverse.

naniar also provides handy visualations for each variable:

 
gg_miss_var(airquality)

Or the number of missings in a given variable at a repeating span

gg_miss_span(pedestrian,
             var = hourly_counts,
             span_every = 1500)

You can read about all of the visualisations in naniar in the vignette Gallery of missing data visualisations using naniar.

naniar also provides handy helpers for calculating the number, proportion, and percentage of missing and complete observations:

n_miss(airquality)
#> [1] 44
n_complete(airquality)
#> [1] 874
prop_miss(airquality)
#> [1] 0.04793028
prop_complete(airquality)
#> [1] 0.9520697
pct_miss(airquality)
#> [1] 4.793028
pct_complete(airquality)
#> [1] 95.20697

Numerical summaries for missing data

naniar provides numerical summaries of missing data, that follow a consistent rule that uses a syntax begining with miss_. Summaries focussing on variables or a single selected variable, start with miss_var_, and summaries for cases (the initial collected row order of the data), they start with miss_case_. All of these functions that return dataframes also work with dplyr's group_by().

For example, we can look at the number and percent of missings in each case and variable with miss_var_summary(), and miss_case_summary(), which both return output ordered by the number of missing values.

 
miss_var_summary(airquality)
#> # A tibble: 6 x 3
#>   variable n_missing   percent
#>      <chr>     <int>     <dbl>
#> 1    Ozone        37 24.183007
#> 2  Solar.R         7  4.575163
#> 3     Wind         0  0.000000
#> 4     Temp         0  0.000000
#> 5    Month         0  0.000000
#> 6      Day         0  0.000000
miss_case_summary(airquality)
#> # A tibble: 153 x 3
#>     case n_missing  percent
#>    <int>     <int>    <dbl>
#>  1     5         2 33.33333
#>  2    27         2 33.33333
#>  3     6         1 16.66667
#>  4    10         1 16.66667
#>  5    11         1 16.66667
#>  6    25         1 16.66667
#>  7    26         1 16.66667
#>  8    32         1 16.66667
#>  9    33         1 16.66667
#> 10    34         1 16.66667
#> # ... with 143 more rows

You could also group_by() to work out the number of missings in each variable across the levels within it.

 
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
airquality %>%
  group_by(Month) %>%
  miss_var_summary()
#> # A tibble: 25 x 4
#>    Month variable n_missing  percent
#>    <int>    <chr>     <int>    <dbl>
#>  1     5    Ozone         5 16.12903
#>  2     5  Solar.R         4 12.90323
#>  3     5     Wind         0  0.00000
#>  4     5     Temp         0  0.00000
#>  5     5      Day         0  0.00000
#>  6     6    Ozone        21 70.00000
#>  7     6  Solar.R         0  0.00000
#>  8     6     Wind         0  0.00000
#>  9     6     Temp         0  0.00000
#> 10     6      Day         0  0.00000
#> # ... with 15 more rows

You can read more about all of these functions in the vignette "Getting Started with naniar".

Contributions

Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.

Future Work

  • Extend the geom_miss_* family to include categorical variables, Bivariate plots: scatterplots, density overlays
  • SQL translation for databases
  • Big Data tools (sparklyr, sparklingwater)
  • Work well with other imputation engines / processes
  • Provide tools for assessing goodness of fit for classical approaches of MCAR, MAR, and MNAR (graphical inference from nullabor package)

Acknowledgements

Firstly, thanks to Di Cook for giving the initial inspiration for the package and laying down the rich theory and literature that the work in naniar is built upon. Naming credit (once again!) goes to Miles McBain. Among various other things, Miles also worked out how to overload the missing data and make it work as a geom. Thanks also to Colin Fay for helping me understand tidy evaluation and for features such as replace_to_na, miss_*_cumsum, and more.

A note on the name

naniar was previously named ggmissing and initially provided a ggplot geom and some other visualisations. ggmissing was changed to naniar to reflect the fact that this package is going to be bigger in scope, and is not just related to ggplot2. Specifically, the package is designed to provide a suite of tools for generating visualisations of missing values and imputations, manipulate, and summarise missing data.

Well, I think it is useful to think of missing values in data being like this other dimension, perhaps like C.S. Lewis's Narnia - a different world, hidden away. You go inside, and sometimes it seems like you've spent no time in there but time has passed very quickly, or the opposite. Also, NAniar = na in r, and if you so desire, naniar may sound like "noneoya" in an nz/aussie accent. Full credit to @MilesMcbain for the name, and @Hadley for the rearranged spelling.

News

naniar 0.1.0 (2017/08/09) "The Founding of naniar"

=========================

  • This is the first release of naniar onto CRAN, updates to naniar will happen reasonably regularly after this approximately every 1-2 months

naniar 0.0.9.9995 (2017/08/07)

=========================

Name change

  • After careful consideration, I have changed back to naniar

Major Change

  • three new functions : miss_case_cumsum / miss_var_cumsum / replace_to_na
  • two new visualisations : gg_var_cumsum & gg_case_cumsum

New Feature

  • group_by is now respected by the following functions:
    • miss_case_cumsum()
    • miss_case_summary()
    • miss_case_table()
    • miss_prop_summary()
    • miss_var_cumsum()
    • miss_var_run()
    • miss_var_span()
    • miss_var_summary()
    • miss_var_table()

Minor changes

  • Reviewed documentation for all functions and improved wording, grammar, and style.
  • Converted roxygen to roxygen markdown
  • updated vignettes and readme
  • added a new vignette "naniar-visualisation", to give a quick overview of the visualisations provided with naniar.
  • changed label_missing* to label_miss to be more consistent with the rest of naniar
  • Add pct and prop helpers (#78)
  • removed miss_df_pct - this was literally the same as pct_miss or prop_miss.
  • break larger files into smaller, more manageable files (#83)
  • gg_miss_var gets a show_pct argument to show the percentage of missing values (Thanks Jennifer for the helpful feedback! :))

Minor changes

  • miss_var_summary & miss_case_summary now have consistent output (one was ordered by n_missing, not the other).
  • prevent error in miss_case_pct
  • enquo_x is now x (as adviced by Hadley)
  • Now has ByteCompile to TRUE
  • add Colin to auth

narnia 0.0.9.9400 (2017/07/24)

=========================

new features

  • replace_to_na is a complement to tidyr::replace_na and replaces a specified value from a variable to NA.
  • gg_miss_fct returns a heatmap of the number of missings per variable for each level of a factor. This feature was very kindly contributed by Colin Fay.
  • gg_miss_ functions now return a ggplot object, which behave as such. gg_miss_ basic themes can be overriden with ggplot functions. This fix was very kindly contributed by Colin Fay.
  • removed defunct functions as per #63
  • made add_* functions handle bare unqouted names where appropriate as per #61
  • added tests for the add_* family
  • got the svgs generated from vdiffr, thanks @karawoo!

breaking changes

  • changed geom_missing_point() to geom_miss_point(), to keep consistent with the rest of the functions in naniar.

narnia 0.0.8.9100 (2017/06/23)

=========================

new features

  • updated datasets brfss and tao as per #59

narnia 0.0.7.9992 (2017/06/22)

=========================

new features

  • add_label_missings()

  • add_label_shadow()

  • cast_shadow()

  • cast_shadow_shift()

  • cast_shadow_shift_label()

  • added github issue / contribution / pull request guides

  • ts generic functions are now miss_var_span and miss_var_run, and gg_miss_span and work on data.frame's, as opposed to just ts objects.

  • add_shadow_shift() adds a column of shadow_shifted values to the current dataframe, adding "_shift" as a suffix

  • cast_shadow() - acts like bind_shadow() but allows for specifying which columns to add

  • shadow_shift now has a method for factors - powered by forcats::fct_explicit_na() #3

bug fixes

  • shadow_shift.numeric works when there is no variance (#37)

name changes

  • changed is_na function to label_na
  • renamed most files to have tidy-miss-[topic]
  • gg_missing_* is changed to gg_miss_* to fit with other syntax

Removed functions

  • Removed old functions miss_cat, shadow_df and shadow_cat, as they are no longer needed, and have been superceded by label_missing_2d, as_shadow, and is_na.

minor changes

  • drastically reduced the size of the pedestrian dataset, consider 4 sensor locations, just for 2016.

New features

  • New dataset, pedestrian - contains hourly counts of pedestrians
  • First pass at time series missing data summaries and plots:
    • miss_ts_run(): return the number of missings / complete in a single run
    • miss_ts_summary(): return the number of missings in a given time period
    • gg_miss_ts(): plot the number of missings in a given time period

Name changes

  • renamed package from naniar to narnia - I had to explain the spelling a few times when I was introducing the package and I realised that I should change the name. Fortunately it isn't on CRAN yet.

naniar 0.0.6.9100 (2017/03/21)

=========================

  • Added prop_miss and the complement prop_complete. Where n_miss returns the number of missing values, prop_miss returns the proportion of missing values. Likewise, prop_complete returns the proportion of complete values.

Defunct functions

  • As stated in 0.0.5.9000, to address Issue #38, I am moving towards the format miss_type_value/fun, because it makes more sense to me when tabbing through functions.

The left hand side functions have been made defunct in favour of the right hand side. - percent_missing_case() --> miss_case_pct() - percent_missing_var() --> miss_var_pct() - percent_missing_df() --> miss_df_pct() - summary_missing_case() --> miss_case_summary() - summary_missing_var() --> miss_var_summary() - table_missing_case() --> miss_case_table() - table_missing_var() --> miss_var_table()

naniar 0.0.5.9000 (2016/01/08)

=========================

Deprecated functions

  • To address Issue #38, I am moving towards the format miss_type_value/fun, because it makes more sense to me when tabbing through functions.
  • miss_* = I want to explore missing values
  • miss_case_* = I want to explore missing cases
  • miss_case_pct = I want to find the percentage of cases containing a missing value
  • miss_case_summary = I want to find the number / percentage of missings in each case miss_case_table = I want a tabulation of the number / percentage of cases missing

This is more consistent and easier to reason with.

Thus, I have renamed the following functions: - percent_missing_case() --> miss_case_pct() - percent_missing_var() --> miss_var_pct() - percent_missing_df() --> miss_df_pct() - summary_missing_case() --> miss_case_summary() - summary_missing_var() --> miss_var_summary() - table_missing_case() --> miss_case_table() - table_missing_var() --> miss_var_table()

These will be made defunct in the next release, 0.0.6.9000 ("The Wood Between Worlds").

naniar 0.0.4.9000 (2016/12/31)

=========================

New features

  • n_complete is a complement to n_miss, and counts the number of complete values in a vector, matrix, or dataframe.

Bug fixes

  • shadow_shift now handles cases where there is only 1 complete value in a vector.

Other changes

  • added much more comprehensive testing with testthat.

naniar 0.0.3.9901 (2016/12/18)

=========================

After a burst of effort on this package I have done some refactoring and thought hard about where this package is going to go. This meant that I had to make the decision to rename the package from ggmissing to naniar. The name may strike you as strange but it reflects the fact that there are many changes happening, and that we will be working on creating a nice utopia (like Narnia by CS Lewis) that helps us make it easier to work with missing data

New Features (under development)

  • add_n_miss and add_prop_miss are helpers that add columns to a dataframe containing the number and proportion of missing values. An example has been provided to use decision trees to explore missing data structure as in Tierney et al

  • geom_miss_point() now supports transparency, thanks to @seasmith (Luke Smith)

  • more shadows. These are mainly around bind_shadow and gather_shadow, which are helper functions to assist with creating

Bug fixes

  • geom_missing_point() broke after the new release of ggplot2 2.2.0, but this is now fixed by ensuring that it inherits from GeomPoint, rather than just a new Geom. Thanks to Mitchell O'hara-Wild for his help with this.

  • missing data summaries table_missing_var and table_missing_case also now return more sensible numbers and variable names. It is possible these function names will change in the future, as these are kind of verbose.

  • semantic versioning was incorrectly entered in the DESCRIPTION file as 0.2.9000, so I changed it to 0.0.2.9000, and then to 0.0.3.9000 now to indicate the new changes, hopefully this won't come back to bite me later. I think I accidentally did this with visdat at some point as well. Live and learn.

Other changes

  • gathered related functions into single R files rather than leaving them in their own.

  • correctly imported the %>% operator from magrittr, and removed a lot of chaff around @importFrom - really don't need to use @importFrom that often.

ggmissing 0.0.2.9000 (2016/07/29)

=========================

New Feature (under development)

  • geom_missing_point() now works in a way that we expect! Thanks to Miles McBain for working out how to get this to work.

ggmissing 0.0.1.9000 (2016/07/29)

=========================

New Feature (under development)

  • tidy summaries for missing data:
    • percent_missing_df returns the percentage of missing data for a data.frame
    • percent_missing_var the percentage of variables that contain missing values
    • percent_missing_case the percentage of cases that contain missing values.
    • table_missing_var table of missing information for variables
    • table_missing_case table of missing information for cases
    • summary_missing_var summary of missing information for variables (counts, percentages)
    • summary_missing_case summary of missing information for variables (counts, percentages)
  • gg_missing_col: plot the missingness in each variable
  • gg_missing_row: plot the missingness in each case
  • gg_missing_which: plot which columns contain missing data.

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("naniar")

0.3.1 by Nicholas Tierney, 17 days ago


https://github.com/njtierney/naniar


Report a bug at https://github.com/njtierney/naniar/issues


Browse source code at https://github.com/cran/naniar


Authors: Nicholas Tierney [aut, cre] (<https://orcid.org/0000-0003-1460-8722>), Di Cook [aut] (<https://orcid.org/0000-0002-3813-7155>), Miles McBain [aut] (<https://orcid.org/0000-0003-2865-2548>), Colin Fay [aut] (<https://orcid.org/0000-0001-7343-1846>), Jim Hester [ctb], Luke Smith [ctb]


Documentation:   PDF Manual  


MIT + file LICENSE license


Imports dplyr, ggplot2, purrr, tidyr, tibble, magrittr, stats, visdat, rlang, forcats, viridis, glue

Suggests knitr, rmarkdown, testthat, rpart, rpart.plot, covr, gridExtra, wakefield, vdiffr, here, simputation, imputeTS, UpSetR, gdtools


See at CRAN