This is a patch release that fixes crashes and incorrect results in group summaries, frequency and cross tables, and summary statistics.
ds_group_summary(), ds_group_summary_interact(),
ds_summary_stats(), ds_freq_table(), ds_measures_variation() and
ds_tidy_stats(), failed with an error on grouped input.ds_group_summary() no longer fails when the continuous variable contains
missing values; groups are summarised with na.rm = TRUE and the number of
missing observations is reported.ds_group_summary() no longer fails when the grouping factor has unused
levels; unused levels are dropped.ds_group_summary() drops observations with a missing grouping label, so the
tidy statistics match the summary table.ds_group_summary_interact() no longer truncates factor levels containing
., and character grouping columns are now used instead of being silently
dropped.print.ds_group_summary() titles the table with the variables, e.g.
mpg by cyl, instead of an empty by.ds_measures_variation(), ds_measures_symmetry(), ds_percentiles() and
ds_tidy_stats() now work on data frames with a single numeric column.ds_measures_variation(), ds_measures_symmetry(), ds_percentiles() and
ds_tidy_stats() handle missing values per column instead of deleting rows
listwise, so missing values in one variable no longer change results for
another. A column with no non-missing values returns NA for every summary
metric instead of -Inf/Inf or an error
(#132).ds_measures_variation(), ds_measures_symmetry() and ds_percentiles()
now apply decimals to data-frame input, as they already did for numeric
vectors.ds_measures_location(), ds_measures_variation(),
ds_measures_symmetry() and ds_percentiles() stop with a clear error when
a non-numeric column is selected explicitly; with no selection they use the
numeric columns
(#152).ds_measures_location() keeps the column order selected by the user, and the
data's column order when no columns are selected, instead of sorting the rows
alphabetically
(#152).ds_summary_stats() and ds_extreme_obs() now work when a variable has fewer
than five non-missing values.ds_summary_stats() validates every selected column before printing, so a
mixed selection fails up front and names the offending column.ds_extreme_obs() and ds_rindex() report the original row indices of the
data supplied, report ties on distinct rows, and ds_extreme_obs() now
applies decimals to data-frame input
(#133).ds_tailobs() checks n against the number of non-missing observations, so
it no longer returns NA values when n exceeds that number.ds_gmean() computes on the log scale, avoiding overflow for large values and
long vectors, and ds_gmean()/ds_hmean() return NA with a warning for
non-positive values
(#134).ds_freq_table() no longer double-counts values that fall on a bin boundary:
intervals are half-open with the last interval closed, so frequencies sum to
the number of observations and, for complete data, the final cumulative
percent is 100 (missing values stay in the percentage denominator). It also
gains a closed argument ("left" default, "right", or "both" for the
legacy double-counting behaviour)
(#131).bins must be a positive whole number; values such as 0, 2.7 and NA are
rejected with a clear error instead of a base error or silent truncation.utility$missing and print Missing Observations when there are any;
ds_auto_cross_table() prints each table's own observation count instead of
the number of rows in the data
(#152).ds_twoway_table() no longer prints a dplyr join message.ds_auto_cross_table(), ds_plot_scatter(), ds_plot_bar_stacked() and
ds_plot_bar_grouped() count usable variables, also when no columns are
selected, and stop with a clear message when fewer than two are available
instead of failing with a low-level error
(#152).factor()) by
the frequency tables, cross tables, group summaries, automatic tables and
categorical plots, so results match the same data stored as factors
(#152).ds_auto_summary_stats() no longer prints a stray NULL after each summary.ds_freq_cont() forwards its variable correctly and no longer
fails with object 'mpg' not found.ds_group_summary() and ds_group_summary_interact() no longer overwrite
data columns whose names match a statistic, such as missing, min or
Levels, so the printed table and tidy_stats report the same values.ds_group_summary() stops with a clear message when the grouping variable
has no non-missing values, instead of failing with
non-numeric argument to mathematical function.ds_summary_stats() and ds_summary() report missing metrics, instead of
Inf/-Inf/NaN, for a variable with no non-missing observations, and
ds_freq_table() stops with a clear message for continuous data with no
non-missing observations instead of printing Inf/NaN bins.ds_hmean() scales the reciprocals before summing, so it returns the value
itself for identical very small positive numbers instead of 0. It also
warns and returns NA when infinite values remain after missing values
(including NaN) are removed.max(x) - min(x) overflows, so a finite range such as
-1e308 to 1e308 produces finite, increasing boundaries instead of
Inf.ds_tailobs() requires n to be a single whole number, returns
numeric(0) for n = 0 and rejects negative, fractional, missing or
infinite values instead of returning a wrong number of observations.ds_twoway_table() derives its join keys from the selected columns, so
programmatic calls with !!rlang::sym() or {{ }} work again.ds_group_summary_interact() groups by the values of the selected columns
rather than a joined label, so label combinations containing . (and any
other separator) are no longer merged.ds_group_summary() and ds_group_summary_interact() return unique column
names. A grouping column named like a statistic, such as length or mean,
is suffixed with _group, so every column keeps its values and the result
can be used with dplyr verbs.ds_group_summary()$tidy_stats keeps the class of the grouping variable,
including ordered factors, and the order of the observed levels.ds_group_summary()$tidy_stats stores length as a double, and quartiles
returned by ds_group_summary_interact() no longer carry the 25%/75%
names inherited from quantile().ds_measures_variation(), ds_measures_symmetry(), ds_percentiles() and
ds_std_error() handle missing values consistently for numeric vectors:
n counts the non-missing observations, measures ignore missing values, and
a vector with no usable values returns NA instead of raising a base error.Missing row and the level/bin percentages share one denominator and add up
to 100.ds_auto_freq_table() no longer triggers the tidyselect deprecation warning
about external vectors.ds_summary_stats() can handle both data.frame and numeric vector as input.This is a patch release to fix CRAN notification about vdiffr package and remove deprecated functions related to visualization of statistical distributions.
The following functions are defunct. Please use the vistributions package for visualizing distributions.
dist_binom_plot()dist_binom_prob()dist_binom_perc()dist_chisquare_plot()dist_chisquare_perc()dist_chisquare_prob()dist_normal_plot()dist_normal_perc()dist_normal_prob()dist_f_plot()dist_f_perc()dist_f_prob()dist_t_plot()dist_t_perc()dist_t_prob()This is a patch release to fix CRAN errors.
ds_tidy_stats() will detect continuous variables in the data set and return
summary statistics for all of them. (#68)ds_auto_summary_stats() will detect all continuous variables in the data
set and return summary statistics and frequency tables for all of them.
(#69)ds_plot_*() family of functions will detect continuous and categorical
variables in the data set and generate appropriate plots. (#70)ds_launch_shiny_app() will check if suggested packages are available and
offer to install the missing packages. (#76)ds_measures_*() family of functions now accept multiple arguments. (#78)ds_auto_freq_table() and ds_auto_cross_table() allow users to specify a
subset of variables. (#83)ds_freq_table() now works for both continuous and categorical data. (#90)All the dist_* functions related to visualizing probability distributions have
been soft deprecated and will be removed in the next release. Please use the
vistributions package going forward.
The shiny app has been soft-deprecated and will be removed in the next release. Please use the xplorerr package going forward.
The following functions have been soft-deprecated and will be removed in the next release:
ds_freq_cont()ds_oway_tables()ds_tway_tables()ds_freq_cont() should return a tibble
(#34)The following now return tibbles to facilitate further analysis:
ds_twoway_table()ds_freq_table()ds_freq_cont()ds_group_summary()New to descriptr are the following which generate summary statistics:
ds_measures_location()ds_measures_variation()ds_measures_symmetry()ds_percentiles()ds_extreme_obs()descriptr:::print_screen() fails if a variable has more than one class
(#26)ds_summary_stats() does not show missing values
(#30)A big thanks to @GegznaV and @adam_medcalf who contributed code, opened issues and provided valuable feedback.
We have completely revamped the API. All the functions now
take a data.frame or tibble as the first argument followed
by the variable names. The variable names need not be surrounded
by single/double quotes anymore. Please view the guide for
more details.
This is a minor release containing bug fixes.
group_summary(); maximum and minimum were reversednorm_plot() was non-reactive to changes in standard deviation (#1).