New ggrocplot() draws a publication-ready ROC curve with the area under the
curve (AUC) and its confidence interval computed and annotated on the plot;
passing several predictor columns overlays one curve per model for comparison,
with each AUC shown in the legend. The empirical ROC, the AUC (the Mann-Whitney
statistic) and a deterministic large-sample confidence interval (Hanley and
McNeil, 1982) are all computed in base R, so the function has no modeling
dependency and runs in lightweight environments such as WebR. For DeLong or
bootstrap intervals, ROC smoothing, or tests between curves, use the pROC
package. When a predictor's AUC falls below 0.5 (higher values indicate the
negative class), a message hints that the predictor or the response levels may
need reversing; the curve is never flipped automatically. Set youden = TRUE
to mark the optimal cut-point (the operating point maximizing the Youden index,
sensitivity + specificity - 1) on each curve and label its threshold
(#766, #767, #768).
New ggcompare() draws a publication-ready group comparison figure in one
call: a box plot (or violin/stripchart) with jittered points and means,
pairwise comparison brackets with adjusted p-values (packed compactly), and an
omnibus test subtitle. It composes existing ggpubr layers (geom_pwc(),
add_test_label()) into one customizable ggplot; every piece can be turned on
or off. It follows the ggfunc contract, so ggsummarystats(ggfunc = ggcompare) places a summary table under the comparison figure.
Mapping a second factor through color or fill composes a two-way (grouped)
comparison figure: dodged boxes for the two factors, simple pairwise comparison
brackets of one factor within each level of the other (pwc.group.by), and a
subtitle from the two-way ANOVA that names the interaction (e.g.
"Interaction (gender × education_level), F(2,52) = 7.34, p = 0.002"). In this
mode the pairwise method defaults to "emmeans_test", p.adjust.method to
"bonferroni" and hide.ns to TRUE; each remains overridable.
Two-way mode supports facet.by: the simple pairwise comparisons and the
two-way interaction are then computed within each facet panel, and a compact
interaction label is drawn in each panel in place of the single pooled
subtitle. It also accepts an opt-in simple.effects = TRUE, which draws each
x group's simple main effect of the compared factor above its brackets,
computed with the pooled error term from the full two-way model so the
denominator degrees of freedom are the full-model residual df
(#759, #763, #769, #772).
New ggvolcano() draws a publication-ready volcano plot of differential
expression results: log2 fold change versus -log10 of the (adjusted)
p-value, with points colored as up/down/non-significant by fold-change and
significance thresholds, the top hits labeled, and threshold lines. It mirrors
the ggmaplot() surface (fdr, fc, top, select.top.method — including
"rank.sum" — top.balanced, label.select, facet.by) and accepts custom
fold-change/p-value column names via x/y (#761).
New ggestimates() draws a publication-ready forest plot: one row per group
with a point estimate, its confidence interval, a reference (null) line and an
optional right-hand column of the numeric estimate and interval. It is a single
point-and-interval builder covering both forest plots (odds/hazard/risk ratios,
regression coefficients, meta-analysis) and estimation plots of effect sizes.
It draws pre-computed values (e.g. from broom::tidy(model, conf.int = TRUE)),
with ref.line, log.scale for ratio estimates, group coloring, row sorting
and an optional point-size (weight) mapping. The default follows the forest-plot
convention (no vertical axis line, right-aligned row labels); set
banding = TRUE for shaded alternate rows (#765).
New ggraincloud() draws a raincloud plot: a half violin (the "cloud"), a
narrow box plot, and the raw jittered data points (the "rain") offset to the
opposite side. Supports coloring by group, journal palettes, horizontal
(default) or vertical orientation, and faceting. The half violin is provided
by the new geom_violin_half() / GeomViolinHalf, a one-sided
geom_violin() usable on its own (#760).
New stat_cld() adds a compact letter display (CLD) of all-pairwise
comparisons to a plot: one letter per group, placed above each box/violin,
where groups that do not share a letter are significantly different. Letters
are derived with rstatix::add_cld() from an all-pairwise post-hoc test
(method: "tukey_hsd" (default), "games_howell_test", "dunn_test",
"wilcox_test" or "t_test"). This is a compact alternative to drawing many
significance brackets when comparing several groups (#755).
New add_test_label() adds, in one call, the omnibus test result (one-way
ANOVA or Kruskal-Wallis) as a plot subtitle, and optionally a pairwise
comparison description as a caption (caption = TRUE). It reproduces the
common + labs(subtitle = get_test_label(...), caption = get_pwc_label(...))
idiom, reusing the re-exported rstatix label helpers. group.by and effect
label a two-way ANOVA effect: with group.by set, the subtitle reports the
two-way y ~ x * group.by model, and effect selects the interaction
(default), a named term, or "all" for a multi-line label of both main effects
and the interaction (#758, #763).
New style_scatterhist() and style_summarystats() restyle the sub-plots of a
ggscatterhist() or ggsummarystats() composite in one call (e.g.
style_scatterhist(p, main = theme_bw(), margin = theme_void())), returning the
same compound object. The documentation now describes the three ways to theme
these composites: at build time (ggtheme=, margin.ggtheme=), by editing the
sub-plots ($sp/$xplot/$yplot, $main.plot/$summary.plot), or via the new
helpers. These composites are lists of plots, not single ggplots, so + does not
restyle them; use one of those options instead (#764).
Re-exported the rstatix annotation helpers get_test_label(),
get_pwc_label(), create_test_label(), add_cld(), and
add_xy_position(), so they can be used directly from ggpubr without
attaching rstatix (e.g. ggpubr::add_xy_position()). These track rstatix's
signatures (#754).
geom_pwc() and stat_pwc() gain a {effsize} label token to display the
pairwise effect size on each bracket, e.g.
label = "{p.adj.signif}, d={effsize}". The effect size is the one returned
by the chosen method (Cohen's d for "t_test"/"games_howell_test",
Cliff's delta for "wilcox_test", r for "dunn_test"), rounded to two
decimals. Labels without {effsize} are unchanged (#756).
geom_pwc() and stat_pwc() also gain a pack argument. With
pack = "auto", pairwise comparison brackets are packed onto the fewest
possible levels so that comparisons whose x-spans do not overlap share a level,
keeping the annotation compact while guaranteeing that no two brackets on the
same level overlap within a panel. The default pack = "none" keeps the
previous one-bracket-per-level stacking, so existing plots are unchanged
(#757).
ggmaplot() and ggvolcano() gain two opt-in options for choosing which
genes are labeled: select.top.method = "rank.sum" ranks by the sum of the
|log2FC| rank and the adjusted-p rank, so the labeled genes are both
high-effect and highly significant; and top.balanced = TRUE labels about
half up-regulated and half down-regulated genes (with spillover). The default
selection is unchanged (#762).
Five vignettes, where the previous release shipped none: "Adding Statistical
Test Results to Plots" (pairwise brackets, compact letter display and omnibus
labels, and when to use each), "Two-Way Comparison Figures with ggcompare()"
(including a section on adding per-group simple main effects, computed with the
pooled error from the full two-way model, anova_test(..., error = model)),
"Forest and Estimation Plots with ggestimates()" (from a table or a fitted
model), "ROC Curves and Diagnostic Accuracy with ggrocplot()", and "Combining
ggpubr Plots with patchwork" (shared legends, panel tags, nested layouts;
points to ggcorrplot for correlation matrices) (#770, #771).
Function help pages and the README now link to the corresponding Datanovia tutorials.
ggsummarystats(free.panels = TRUE) now rejects an invalid labeller with a
message naming the argument, instead of failing with whatever the internal
lookup happened to raise. Both documented values, "label_value" and
"label_both", behave exactly as before. Values outside them are now refused
up front: a number, TRUE, or a factor used to be accepted and picked a
labeller by position, because the lookup was a switch() that indexes by
integer when its argument is not a character string — so TRUE and
factor("label_value") both selected "label_both", the opposite of the
request in the second case, and 2 selected "label_value".
free.panels = FALSE already refused a number and TRUE, though it still
accepts a factor (#781).
Naming the same item in both select = and remove = now warns, listing the
items in common. The call asks to keep and drop the same item; remove is
applied after select, so the item is dropped. Nothing about the drawn result
changes — the warning only makes the resolution visible, since the call
previously went through silently. Affects the eight builders that declare both
arguments (ggboxplot(), ggviolin(), ggbarplot(), ggline(),
ggstripchart(), ggdotplot(), ggdotchart(), ggerrorplot()) and the six
that accept them through ... (ggscatter(), ggdensity(), gghistogram(),
ggecdf(), ggqqplot(), ggpaired()) (#788).
rstatix version is now >= 1.1.0 (#753).ggcompare(), ggrocplot(), ggestimates(),
ggraincloud(), ggvolcano(), stat_cld(), add_test_label(),
geom_violin_half(), style_scatterhist(), style_summarystats()), and the
README highlights them (#773).ggsummarystats() now draws its summary table on the categories the plot
was drawn on, so each column sits under the box it describes. The table
trained its own x scale, so when the two orders differed the columns were
shuffled relative to the boxes: with time = Pre/Post/Mid the column under
Pre printed Mid's median. The columns were shuffled by order = in the
same way, and select = / remove = left behind a column for a group the
plot no longer drew. A POSIXct x, which could not be printed at all, now
draws. $summary.plot$data holds only the groups drawn, with its x column a
factor over those categories. Plots whose x axis is continuous — ggline()
or ggbarplot() with an integer or Date x, or any builder called with
numeric.x.axis = TRUE — and facets with scales = "free_x" are not
affected by this change. See ?ggsummarystats (#780).
ggsummarystats(free.panels = TRUE) now titles every panel with the group
whose data it draws. The label was built from a sorted copy of the rows and
attached to the unsorted data, so panels were titled with each other's names: a
panel headed East drew North's boxes, and the summary table under it
carried North's n, median and IQR. The label is now built from the grouping
key that travels with each panel's own rows, so it cannot drift.
This affected any facet.by column whose groups do not already appear in sorted
order: a character column that is not alphabetical, a factor whose levels are
ordered differently from its rows, numeric, integer, logical and Date columns,
and a frame whose NA group is not last. A column whose groups happen to appear
in sorted order was unaffected. labeller = "label_both" was hit more often,
because it sorts the formatted variable:value strings rather than the
underlying values, so it also transposed factor and numeric columns that
label_value got right. A two-variable case such as
facet.by = c("supp", "qc") with free.panels = TRUE had all four of its
panels transposed; if you have produced such a figure it is worth re-checking.
Each panel keeps its position and its data; what changes is the name written on
it, so the titles now read in the order the groups appear in the data rather
than in sorted order. Label formatting is otherwise unchanged, except where a
facet column is named label, which the label machinery uses internally: such
a column now keeps the label: prefix that label_both asks for, and keeps
the other facet variables in the label under label_value. Both were dropped
before, so facet.by = c("region", "label") returned two panels named p and
two named q with region missing from all four, and only the first of each
pair could be reached by name. The panel-title defect was reported upstream
(rstatix issue 324) and is fixed in rstatix's development version, so with that
installed the labels were already right; ggpubr requires rstatix (>= 1.1.0),
the released version carrying the defect, and the label-column case above is
wrong on any rstatix version, so the fix is applied here regardless (#779).
ggsummarystats(comparisons = ...) now routes to ggcompare() automatically
and draws the comparison brackets and omnibus test label that were asked for.
Previously comparisons was silently dropped and the plot was drawn without
them (#759).
ggbarplot() now keeps each error bar on its own bar when a variable is mapped
to alpha and the bars are dodged, under both position_dodge() and
position_dodge2(). Thanks to @zuooonz (#404). The error layer dodged by a
different key from the one ggplot2 uses to group the bars, so in a design with
three discrete variables half the error bars were centered on an adjacent bar's
mean and carried its error. Under position_dodge2() most such calls failed at
draw, with "alpha * 255": non-numeric argument to binary operator, because
the summary dropped the alpha column and its name reached the graphics engine
as a static opacity; the ones that did draw had alpha mapped to a column that
was already a grouping variable (for example fill = "supp", alpha = "supp"),
so the summary already carried it, and those were drawn with the error bars on
the wrong bars.
The key is now built from every mapped discrete aesthetic, in the order ggplot2
lays them out, and the pairing is checked against the bars before they are
drawn. This also aligns cases that were misplaced before: fill, color and
alpha on three different columns; an alpha column whose levels are reversed,
unused, ordered or logical; one containing NA; a tibble or grouped input;
shuffled input rows; facet.by = of one or two variables, including panels in
which a level is absent and scales = "free_x" (or "free"), where a panel
missing an x level renumbers the levels it does draw and the error bars were
placed at the un-renumbered positions, sometimes past the edge of the panel
(scales = "fixed" and "free_y" are unaffected); sort.val =,
sort.by.groups = FALSE, orientation = "horizontal"; the error.plot =
variants; preserve = "single"; cells that share a mean; unbalanced, singleton
and zero-variance cells; an add.params$color naming the alpha column; and a
user-set add.params$group naming a column unrelated to the legend, which
summarised the error layer by that column so every interval carried another
cell's statistic. This changes the appearance of such a plot, which was
previously drawn with the error bars permuted.
With error.plot = "crossbar" the crossbars were drawn off-centre and wider
than their bars, and are now centred on the bar at the bar's own width and
still drawn as crossbars; their fill still follows the mapped fill. With
error.plot = "crossbar" and an add.params$color naming a data column whose
name is also a colour (a column called "red", say), the argument used to be
dropped and the outline drawn black; it is now used as the colour, matching
what the other error.plot values already did with the same input.
Unchanged from previous releases: the stacked default; a numeric alpha
column; a grouping column named after one of the statistics desc_statby()
computes (length, min, max, median, mean, iqr, mad, sd, se,
ci, range, cv, var) — the summary's column of that name holds the
computed statistic, so the aesthetic mapped to it follows that statistic rather
than the column, and the calls that already failed still fail (such a column on
x errors for all thirteen; on alpha, mean fails at draw and ci when the
column is character); a numeric, integer or Date column mapped to
color/fill, which ggplot2 does not group the bars by, so the layer draws
more bars than there are dodge positions and no error bar can be matched to a
single bar; and label = TRUE, whose value labels are not moved by this fix and
still dodge on the fill key alone, so with a discrete alpha they are drawn
between the bars, two to a position.
Four configurations keep the released behaviour of refusing to draw under
position_dodge2() with a discrete alpha, rather than placing an interval
beside a bar it was not computed from: a color/fill/alpha column that is
not discrete; label = TRUE (or a character label vector); a raw-data layer
(add = c("mean_se", "jitter"), and likewise point, dotplot, boxplot,
violin), which already sits off its own bar without any alpha; and an
asymmetric summary (median_q1q3, median_hilow), which has no half-width
column for the layer to be rebuilt from.
Under plain position_dodge(), top = remains unsupported alongside a
discrete alpha, and is the one configuration whose broken output changed
rather than staying as released: the error layer is still built for every
summary group rather than for the bars actually kept, so it draws more error
bars than there are bars, some over empty space and some overlapping a kept bar
while carrying another group's statistic — but which stray interval lands where
now differs from previous releases. Under position_dodge2() the error layer is
built from the same truncated summary as the bars, so top = is aligned there.
ggbarplot(position = position_dodge2(reverse = TRUE)) now draws each error
bar on its own bar when the bars are grouped by a single discrete
color/fill column — the ordinary case, with no alpha mapping involved
(#783). Every interval was drawn on the neighbouring bar, carrying that
neighbour's mean and error; the values were right, the pairing was not, and
nothing in the figure showed it. position_dodge2() lays each x out in
descending group order when reverse is set — or any value it treats as true,
such as 1 — while the error layer was ordered ascending, so each row was
matched to the mirror bar. This changes the appearance of such a plot, which was
previously drawn with the intervals transposed; every position_dodge2() call
without reverse is unchanged, as is every other position.
Three reverse = TRUE configurations are still misplaced, all as wrong as in
previous releases. The error layer keys on only the first of color/fill,
so whenever that column does not describe how the bars are grouped, no ordering
of it can pair the intervals:
color and fill naming two different discrete columns;color column that is constant or all-NA beside a real fill grouping;color beside a discrete fill.A non-discrete column mapped to color/fill on its own is unchanged for a
different reason: ggplot2 does not group the bars by such a column at all, so
position_dodge2() does not reverse them and the ascending key stays correct.
A faceted scales = "free_x" keeps a pre-existing offset in a panel that is
missing an x level, unrelated to reverse and present in previous releases
(#784).
select = and remove = used together now draw the groups that were asked
for. The row filter was built once from the unfiltered data and then reused
after select had already shrunk it, so the second filter was applied through
a mask longer than the data it indexed. Rows of the removed group were kept, a
block of NA rows formed a phantom category, and the summaries were computed
over the wrong rows: ggboxplot(ToothGrowth, "dose", "len", select = c("0.5", "1"), remove = "2") drew a median of 7.15 for the dose-0.5 box, where
that group's median is 9.85 — 7.15 is the median of a subgroup the box does not
represent. This affected every builder taking the two arguments —
ggboxplot(), ggviolin(), ggbarplot(), ggline(), ggstripchart(),
ggdotplot(), ggdotchart() and ggerrorplot(), and the six that accept the
two arguments through ...: ggscatter(), ggdensity(), gghistogram(),
ggecdf(), ggqqplot() and ggpaired() — whenever select actually dropped
a group. Using one argument on its own was never affected. A data.table was
the one input that caught the mismatch: it refused the over-long mask and the
call errored rather than drawing, so for that input the combination now works
where it previously stopped.
The filter also no longer goes through subset(), which evaluated its
expression inside the data frame, so a column named select, remove or x
took the place of the argument. This is the one way a single-argument call
changes. What was drawn then depended on that column's contents rather than on
what was asked for, so the call could drop nothing, drop everything, or drop
the wrong group — a frame with a column named remove holding "a" answered
remove = "c" by dropping a and keeping c, which looks like an ordinary
figure with the wrong group missing (#788).
ggsummarystats(facet.by = character(0) or "", free.panels = TRUE, labeller = "label_both") no longer errors. A facet.by that resolves to no
grouping column yields a single panel; there is no variable name to build
"<variable>:<value>" from, so the panel is titled "", which is what the
same call has always produced with the default "label_value" (#781).
stat_compare_means(label = "p.format") (and label = "p") no longer fail
with could not find function "create_p_label" when ggpubr is called via
ggpubr:: without being attached by library(ggpubr). The label helper is now
namespace-qualified inside after_stat() so it resolves off the search path.
Reported by @pwwang (#751).
format_p_value(), get_p_format_style(), and list_p_format_styles().default, apa, nejm, lancet, ama, graphpad, scientific).ggmaplot() gains an optional facet.by argument to split the MA plot into
multiple panels (e.g. one per contrast or species). Top genes are selected per
panel, while point colors use the same significance thresholds in every panel.
Default output (no facet.by) is unchanged (#498).
stat_cor() and stat_regline_equation() gain a label.anchor argument.
With label.anchor = "panel" the label is placed at the true panel-relative
position (npc), so labels stay aligned across panels or facets whose axis
ranges differ - e.g. facet_wrap(scales = "free_y"), geom_smooth()
extending each panel by a different amount, or separate plots combined with
ggarrange(). The default label.anchor = "data" keeps the previous
data-range placement, so existing plots are unchanged (#248).
geom_bracket() gains an orientation argument. orientation = "vertical"
draws a native vertical bracket (the bar spans the y axis, the tips point left
and the label is rotated), specified with ymin/ymax/x.position — useful
to annotate plots where the comparison runs along the y axis, e.g.
Kaplan-Meier curves (it is designed for a continuous x axis). The default
orientation = "horizontal" is unchanged, and it cannot be combined with
coord.flip = TRUE (#456).
ggpaired() gains an opt-in jitter argument that spreads the paired points
horizontally to reduce overlap. Each subject (id) gets a single offset, so
its two points move together and the connecting line stays intact, and only
the horizontal positions change (the values are never moved). Default
jitter = 0 leaves the plot unchanged (#407).
ggboxplot(), ggviolin() and ggstripchart() gain an opt-in show.n
argument. When show.n = TRUE, the number of observations ("n = <count>")
is displayed at the top of each group. When the groups are dodged (a
color/fill grouping with a dodging position), one count is shown per
group; otherwise a single count is shown per x-axis tick. Counts respect
select/remove and are computed per facet. Default is show.n = FALSE, so
existing plots are unchanged (#598, #627).
ggarrange() gains a spacing argument to increase the gap between the
arranged plots (a long-requested option). It adds a uniform margin, in
text-line units, around each plot. Default is 0 (no extra space; each plot
keeps its own margins, so existing arrangements are unchanged) (#151).
compare_means() gains an id argument for paired comparisons. Without it, a
paired test (paired = TRUE) pairs observations by row order, so the p-value is
wrong when the data are not sorted so that the compared groups align by subject.
Passing id (the name of a subject-identifier column) pairs the observations by
id instead — row-order independent, using only the complete pairs (via
rstatix, so the result matches a direct rstatix::t_test()/wilcox_test()
call). It works for a two-group, a pairwise (more than two groups) and a
ref.group comparison with t.test/wilcox.test. Default (id = NULL) is
unchanged (#560).
stat_compare_means() also gains the id argument, so a paired test can be
aligned by subject id directly on a plot (e.g. ggpaired(...) + stat_compare_means(paired = TRUE, id = "subject")), displaying the correct
p-value even when the data are not sorted by subject. Plots without id are
unchanged (#560).
geom_pwc() and stat_pwc() gain a p.adjust.n argument giving the number of
comparisons to use for the p-value adjustment (passed as n to
stats::p.adjust()). Default is NULL (adjust by the number of computed
p-values, unchanged behavior); set it when the displayed comparisons are a
subset of a larger family so the adjustment reflects that larger size. It must
be at least the number of p-values being adjusted, and applies to the
panel-level adjustment (p.adjust.by = "panel") and the single-comparison
case (#612).
annotate_figure() gains column.titles and row.titles arguments to add a
title above each column and to the left of each row of an arranged grid
(useful for publication panels). Each accepts a character vector (one title
per column / per row, rendered in bold) or a list of grobs (e.g. built with
text_grob()) for full styling control; row.titles are rotated 90 degrees.
The titles are placed in reserved space (they do not overlap the plots) and
are evenly spaced, so they assume equal-sized columns / rows (the ggarrange()
default). Default is NULL (no titles; existing figures are unchanged) (#573).
ggbarplot() gains a numeric.x.axis argument (already available in
ggline() and ggerrorplot()). With numeric.x.axis = TRUE the x variable
is kept numeric instead of being coerced to a discrete factor, so bars are
drawn at their numeric x positions (e.g. a time axis). Default is FALSE
(unchanged behavior) (#463).
geom_bracket() and stat_pvalue_manual() gain a tip.length.ref argument
controlling what tip.length is a fraction of. With the default
tip.length.ref = "data" the tips are a fraction of the data range (unchanged
behavior). With tip.length.ref = "axis" the tips are a fraction of the
y-axis range (ylim/scale_y_*), which renders at the same physical fraction
across plots and therefore gives visually constant tip lengths regardless of
the data range - useful to keep tips consistent across facets or across
separate plots with different scales (#362). stat_compare_means(comparisons =)
draws its brackets via ggsignif::geom_signif() and does not support
tip.length.ref; passing it there now emits an informative message pointing to
stat_pvalue_manual(..., tip.length.ref = "axis") instead of being silently
ignored (#362).
stat_cor() and stat_regline_equation() gain a label.y.step argument giving
the vertical spacing (in text-line units) between the labels of successive
groups. Default is 1.4 (unchanged behavior). Setting label.y.step = 0 stops
the per-group vertical shift so labels align across facet panels when a factor
is mapped to an aesthetic that also defines the facets (the long-standing
"labels climbing stairs across facets" behavior; #248). This control was
inspired by the vstep argument in the ggpmisc package (by Pedro J. Aphalo),
from which ggpubr's correlation/regression label-positioning logic was
originally adapted; this is acknowledged via @seealso in ?stat_cor.
ggviolin() gains a drop argument (passed to ggplot2::geom_violin()),
default TRUE (unchanged behavior). Previously the argument was silently
dropped, so grouped violins where a sub-sample has too few points to draw a
density (which geom_violin() removes, including from the dodge position)
could not be kept aligned with added boxplots / dot plots. Setting
drop = FALSE together with position = position_dodge(0.8, preserve = "single")
now reserves the empty dodge lane so all geoms stay aligned (#381).
ggarrange() gained the ability to choose which plot(s) supply the shared legend:
common.legend now also accepts one or several plot indices. common.legend = 2
uses the second plot's legend (handy when the first plot's legend is not
representative, e.g. a group is missing in the first plot), and
common.legend = c(1, 2) keeps and combines the legends of the listed plots into a
single shared block (side by side for legend = "top"/"bottom", stacked for
"left"/"right") - useful when the arranged plots genuinely need different
legends. The documentation of common.legend was also clarified: common.legend = TRUE keeps the first plot's legend (it does not merge or validate legends), so plots
should share a consistent scale for a single shared legend to be correct. Logical
TRUE/FALSE behave exactly as before (#347).
ggdonutchart() gained a label.repel argument (default FALSE). When TRUE, slice
labels are placed with ggrepel::geom_text_repel() and connected to their slice with
leader lines, so the labels of many small slices no longer overlap (or get dropped). The
default output is unchanged (#655).
ggboxplot() now accepts a position argument (e.g.
position = position_dodge(0.9)), like ggviolin()/ggdotplot() already do.
Previously passing position errored (formal argument "position" matched by multiple actual arguments) because the dodge was hardcoded; the default is
unchanged (position_dodge(0.8)), so grouped box plots look the same unless
you set it (#615).
ggsummarytable() gains an angle argument to rotate the summary-table text;
default (angle = 0) is unchanged (#595).
ggmaplot() gains a line.color argument to set the threshold line color;
default ("black") is unchanged (#322).
ggarrange() gains a byrow argument (default TRUE) to fill the plot grid
by column (byrow = FALSE) instead of by row; forwarded to
cowplot::plot_grid() (#225).
stat_regline_equation() gains coef.digits and rr.digits arguments to
control the number of significant digits shown for the regression-equation
coefficients and R2; defaults (2) reproduce the previous output (#312).
Added formatting parameters to statistical helpers:
p.format.stylep.digitsp.leading.zerop.min.thresholdp.decimal.markAdded p.format.signif support and related label handling paths.
stat_pvalue_manual() gains a p.digits argument (default 3) that formats
numeric p-value label columns for display (e.g. label = "p" or
label = "p.adj"), using the same format_p_value() engine as
stat_anova_test(). This restores clean labels (e.g. 0.0156 instead of
0.015625) with rstatix >= 1.0.0, which now returns full-precision pairwise
p-values. Character labels (significance symbols, pre-formatted strings, glue
expressions) are unaffected. Set p.digits = NULL to print the raw value.
stat_pvalue_manual() additionally gains p.format.style, p.leading.zero,
p.min.threshold and p.decimal.mark, matching the p-value formatting
arguments already available in stat_compare_means(), geom_pwc() and
stat_anova_test(). These are opt-in and default so that the rendered labels
are unchanged; for example p.min.threshold = 0.001 displays very small
p-values as < 0.001, and p.format.style = "nejm" applies a journal style.
stat_cor() gains two label-formatting arguments:
r.leading.zero — set to FALSE to drop the leading zero of the
correlation coefficient (e.g. .73 instead of 0.73), completing
APA-style reporting together with p.leading.zero (#540). The dropped
leading zero is preserved through plotmath rendering (the value is quoted
so it is not silently re-normalized back to 0.73 in the default
expression output).p.coef.name — symbol for the p-value label; use "P" for an uppercase
p-value (#541).stat_cor() exposes two new computed variables, rmse and rmse.label, for
the root mean square deviation (RMSE/RMSD) between x and y — useful for
reporting agreement between paired measurements on the same scale (e.g.
predicted vs. reference values). Display it with
aes(label = after_stat(rmse.label)), or combine it with the correlation
coefficient using paste(). The default label is unchanged (#458).
stat_cor() can now display the confidence interval of the correlation
coefficient, via the new computed variables conf.int.low, conf.int.high
and conf.int.label (e.g. "95% CI [0.21, 0.75]") and a conf.level argument
(default 0.95). Show it with aes(label = after_stat(conf.int.label)), or
combine it with the coefficient using paste(). The confidence interval is
available for method = "pearson" only; it is NA for Spearman/Kendall. The
default label is unchanged (#418).
ggplot2, dplyr, and tidyr.size usage to linewidth where required by recent ggplot2.size -> linewidth migration for the remaining line layers that
still passed the deprecated size argument: the mean/median reference line added
by gghistogram(add = ...) and ggdensity(add = ...), and the connector segments
of ggdotchart(add = "segments"). These no longer emit ggplot2's "size aesthetic
for lines was deprecated" or "Ignoring empty aesthetic: size" warnings, and the
requested line width is now applied via linewidth.ggmaplot() no longer emits an "Ignoring empty aesthetic: size" warning on its
default call; the point layer sets size only when the user supplies a value.clean_lock_files() helper.ggrepel dependency to >= 0.9.2 to keep Ubuntu oldrel
(R 4.4.x) CI dependency resolution working.geom_pwc() now detects an absent ref.group in a grouped subset via the
rstatix_missing_ref_group condition class raised by recent rstatix
(walking the parent chain with rlang::cnd_inherits()), in addition to
matching the error message. This keeps the "skip ref-less subsets" behaviour
working after rstatix made that error message clearer (rstatix #153); older
rstatix versions are still handled via the message fallback.rstatix version is now >= 1.0.0 (the version that
introduced the id argument used by compare_means(id = ), and the
full-precision pairwise p-values relied on by the p-value formatting).stat_compare_means() and geom_pwc().geom_pwc() so users can see which grouped
subsets were skipped and why (e.g., missing ref.group or insufficient levels).ggtheme = NULL to the plotting functions (e.g. ggscatter(),
ggboxplot(), ggline(), gghistogram(), ...) now skips applying a ggpubr
theme, so the plot keeps ggplot2's default theme or the theme set globally with
theme_set(). Previously an explicit ggtheme = NULL was treated like an unset
argument and theme_pubr() was still applied. Calls that omit ggtheme or pass
a specific theme are unchanged (#561).compare_means() now adjusts p-values within each group.by level (and
response) rather than pooling all groups together, so a grouped adjustment
matches filtering to one group and adjusting there. This changes p.adj values
for calls that use group.by; ungrouped calls are unchanged (#200).stat_compare_means(comparisons = ) now emits a one-time message (and the docs
note) clarifying that the displayed pairwise p-values are not adjusted for
multiple comparisons, pointing to geom_pwc() / stat_pvalue_manual() +
compare_means(p.adjust.method = ) for corrected p-values (#293).rstatix::add_cld(), then place
them with geom_text() (see the "Significance letters" section in ?compare_means). Also covers
the per-control letters case (a = differs from control 1, b = differs from control 2) (#464, #434).ggbarplot() (e.g. add = "mean_se") must be faceted with
the facet.by= argument, not by appending + facet_wrap()/+ facet_grid(): the summaries
are pre-computed over facet.by, so a manually added facet pools the bars (and, for stacked
bars, the error bars) across all panels (#739).ggtheme default per
function, the p-value threshold/precision defaults, the stat_compare_means()
label separator (test method name, not correlation coefficient), and the
package Description's list of p-value formatting styles. Thanks to @erdeyl (#749).stat_compare_means() fatal warning/error in sparse grouped subsets with modern tidyr).stat_compare_means(label = "p.format") rounding and formatting control).stat_cor()
confidence interval, the stat_cor() RMSE label, and a manual number of
comparisons for p-value adjustment in geom_pwc()/stat_pwc()..get_brewer_pal() so
"YlOrBr" is recognized correctly.ggexport() to respect verbose = FALSE by suppressing filename
print() output in multi-file raster/vector exports.format_p_value() by validating non-NULL p.min.threshold as a
single positive finite number.create_p_label() to preserve NA values in p.format (returning
NA_character_ instead of stringifying to "p = NA").theme_pubr() now draws the axis tick marks in black with linewidth 0.5,
matching the axis lines; previously the ticks inherited a lighter grey/thinner
style, visibly inconsistent when zoomed (#668).theme_pubr() now sets strip.clip = "off" so the facet strip background
border renders at its full linewidth. With the ggplot2 (>= 3.5.0) default
(strip.clip = "on") the border was clipped to the strip area, cutting the
outer half of the stroke so it looked thinner and misaligned with the panel
when zoomed (follow-up to #668). Note: a facet label wider than its panel now
overflows the strip instead of being truncated; restore clipping with
+ theme(strip.clip = "on") if needed.xticks.by/yticks.by now anchor the axis breaks to round multiples of the
step (e.g. 0, 100, 200) instead of the slightly-negative expanded axis minimum,
which produced odd labels such as -20, 80, 180 on bar plots with
ylim = c(0, 400) (#313).%||% operator from rlang; it is used in geom_bracket() and
geom_pwc() but base R only provides it since R 4.4, so it could be unresolved
on the R (>= 4.1) versions the package supports (#665).stat_welch_anova_test(label = "as_detailed_italic") and
stat_welch_anova_test(label = "as_detailed_expression") now display the numeric
F statistic instead of FALSE. Thanks to @erdeyl (#749).ggbarplot() now supports mapping alpha to a discrete grouping variable together
with a summary (e.g. alpha = clarity, add = "mean_ci", position = position_dodge()).
Previously this errored at draw time ("alpha * 255": non-numeric argument to binary operator). The mean/CI is now computed per subgroup, the bars are faded per the alpha
variable, and the error bars are dodged to stay aligned with their bars. Bar plots
without an alpha grouping variable are unchanged (#404).ggpar() no longer errors on ggsurvplot objects (or any plot whose theme uses
ggtext::element_markdown(), e.g. survminer's risk-table strata labels). Such
markdown label elements are now left intact instead of triggering an "Only elements
of the same class can be merged" error; output for all other plots is unchanged (#382).geom_pwc() with an explicit list of comparisons no longer drops a whole facet/panel
when a single requested pair cannot be tested (e.g. a group that is entirely NA or has
fewer than two observations). The untestable comparison is skipped (with a message) and
the remaining valid comparisons are still drawn (#542).compare_means() and stat_compare_means() now forward extra test options
(e.g. alternative = "greater") to the paired tests aligned by subject id,
which were previously ignored on that path. Grouped and faceted paired-id
comparisons also skip a subset that cannot be tested (fewer than two groups or
no complete pairs) instead of failing the whole result/layer, while still
reporting genuinely ambiguous data (e.g. duplicated ids). Thanks to
@erdeyl (#732).ggbarplot() now places the error bars of a stacked bar chart correctly when the
data mix positive and negative values (e.g. above/below-ground measurements). The
stacked error bars are cumulated per sign, matching position_stack(), so a negative
segment's error bar is drawn on its own side instead of being displaced to the other
side (previously the misplacement also flipped with the factor-level order). Stacked
charts with single-sign data are unchanged (#426).ggviolin() now keeps grouped violins aligned with their added box/dot layers by
default when a sub-group is too sparse for geom_violin() to compute a density
(a single data point). Previously the sparse sub-group was dropped from the dodge,
so the remaining violin re-centered and no longer lined up. When drop/position
are left at their defaults and such a one-point grouped cell is present, the empty
dodge lane is now reserved automatically. Balanced, ungrouped, faceted, and
legitimately-unbalanced plots are unchanged, and an explicit drop/position
still takes precedence (#381).geom_bracket() (and stat_pvalue_manual()) now draw visible tips for a single
bracket placed over a stat-computed y such as geom_bar()/geom_histogram().
Previously the bracket tips collapsed into a flat line because the y-axis range
was trained only on the bracket's single y.position; the tips are now sized
against the fully-trained panel range at draw time, so they render correctly even
with coord_cartesian(ylim = ...). Brackets with an already non-zero y range are
unchanged (#631).ggline() now dodges the jittered points together with the line and error
bars. Previously, in a grouped line plot with add = "jitter" and
position = position_dodge(), only the line and summary (e.g. mean_se) were
dodged while the jittered points stayed centered, so they no longer sat under
their group. The jitter now dodges by the same width. Plots without a
position_dodge() (the default position = "identity") are unchanged (#436).ggbarplot() now lines the error bars up with the bars when
position = position_dodge2() is used. position_dodge2() places elements
according to their own width, so the thin error bars did not sit on the centre
of the wider bars they belong to (most visible with preserve = "single" when
one x group has fewer bars than another). The error bars are now drawn at the
actual bar positions. Other positions (identity, position_dodge(),
position_stack()) are unchanged (#363).geom_bracket() and stat_pvalue_manual() now place brackets correctly on a
transformed y axis (e.g. scale_y_log10()). y.position is given in data units
but was previously used directly in the scale's transformed space, so brackets
landed far off (e.g. at 10^30) and squashed the plot. The bracket y.position
is now run through the scale's transformation; on an untransformed (identity)
scale this is a no-op, so existing plots are unchanged (#342).ggscatter(), ggboxplot(), ggviolin(),
ggline(), ggbarplot(), gghistogram(), ggdensity(), ggdotplot(),
ggstripchart(), ggerrorplot(), ggecdf(), ggdotchart(), ggpaired(),
ggqqplot()) now accept the British spelling colour as an alias for
color. Previously colour was silently ignored (#317).stat_regline_equation() now displays the correct equation for orthogonal
polynomial fits such as formula = y ~ poly(x, 2). Previously the orthogonal
basis coefficients were printed as if they were raw polynomial coefficients,
giving a wrong equation that did not match the fitted curve. For a simple
poly(x, k) term with an intercept, the equation is now computed from an
equivalent raw = TRUE fit (identical curve, R², AIC, BIC). Linear,
raw = TRUE, and I(x^2) formulas are unaffected; no-intercept models and
transformed poly arguments (e.g. poly(log(x), 2)) keep their previous
behavior (#653).ggscatterhist() now aligns the marginal plots with the main scatter plot's
axes. The margins were built from the raw data, so anything that changed the
scatter limits (ellipse = TRUE, position jitter, explicit xlim/ylim)
left the marginal histograms/densities misaligned with the scatter. Each
margin now inherits the scatter's axis range (#220, #420).geom_pwc() now places comparison brackets over the correct groups when an
entire x-axis group has only NA values. Such a group is dropped before the
statistical test runs; previously the surviving groups were renumbered from 1,
which shifted the remaining brackets left. The bracket positions are now
anchored to the discrete x scale so they stay aligned with the plotted groups
(#575).ggballoonplot() now honors user-supplied xlab/ylab instead of always
blanking the axis titles; axis titles are still blank by default (#639).ggpar(legend.title = ) now also titles size and alpha legends, not just
colour/fill/linetype/shape (#412).facet()/panel.labs: a NAMED panel.labs vector is now matched to the data
levels by name, fixing mislabeled panels when the order differed; unnamed
labels still map positionally (#643).stat_compare_means() no longer fails with a "*.npc coord ..." error when a
grouped x value is 0 (or negative); .group_coord() now guards against a
non-positive group index and falls back to the first label coordinate (#594).ggpar()/plot functions now apply xticks.by and yticks.by together; an
internal else if previously dropped xticks.by whenever yticks.by was
also supplied (#333).stat_compare_means() now forwards the family argument to the comparison
bracket labels; it was previously ignored whenever comparisons was set
(#592, #624).ggline() now supports two grouping variables at once (e.g. color and
linetype, or an explicit group plus color); these are combined into a
single interaction group so the right points are connected. Previously this
errored with "the condition has length > 1" (#616, #375).palette vector no longer emits a spurious "No shared levels found
between names(values) ..." warning; the manual color/fill scale is now only
applied to an aesthetic actually mapped in the plot. Unnamed palettes are
unchanged, and a named palette whose names genuinely don't match still warns
(#642).ggviolin() no longer crashes with scales = "free" (or "free_x"/"free_y").
The facet argument scales was being partial-matched to the violin scale
parameter; it is now read with exact matching, so scales controls faceting
and the violin scale keeps its default (#398).xlim/ylim are now honored with orientation = "horizontal" (and
rotate = TRUE). The limits are passed to coord_flip() instead of a separate
coord_cartesian() that was silently replaced, which also removes the
"Coordinate system already present" warning (#646).ggpar() (and the plot functions) now honor legend.direction for all legend
positions; it was previously ignored for legend = "top"/"bottom" (#652).compare_means() no longer errors ("Can't extract columns that don't exist")
when a formula variable name contains a space (e.g. len ~ `spa ced` ); the
backticks that R adds to non-syntactic term names are now stripped before
matching data columns (#385).table_cell_font() and table_cell_bg() can now style an individual header
cell (row = 1), not only body cells; they previously matched only the
core-* grobs and silently did nothing for the header (#535).ggmaplot() now draws the non-significant ("NS") points behind the significant
ones, so the up/down-regulated hits are no longer hidden under the grey NS
cloud; the legend order (Up, Down, NS) is unchanged (#365).ggboxplot() now forwards coef to geom_boxplot(), so coef = 0 can be used
to omit the whiskers; it was previously dropped by geom_exec() (#517).tab_add_title() / tab_add_footnote(): the just argument now actually
positions the text — "center"/"right" anchor the title/footnote across the
table width instead of around a fixed point — and the text is no longer clipped
when it is wider than the table (#302).annotate_figure(): the figure label (fig.lab) now uses a length-independent
horizontal justification, so labels of different lengths keep the same anchor
and captions align across figures; previously a longer label was shifted away
from the corner (#185).reframe(), slice_head(), slice_tail(), across(), and where() functions.linewidth parameter to ggboxplot(), gghistogram(), ggviolin(), and ggdensity() for ggplot2 3.4.0+ compatibility. The size parameter is deprecated for line width in these functions (#644, #645, #654, #656, @erdeyl).adjust parameter to ggviolin() to control bandwidth adjustment for kernel density estimation (#552, @erdeyl).bw and adjust parameters to ggdensity() for bandwidth control (#490, @erdeyl).stat_cor(), stat_compare_means(), and stat_regline_equation() now use after_stat() syntax instead of deprecated ..var.. notation in default_aes (#645, @erdeyl).ggballoonplot() example updated to use guides(size = "none") instead of deprecated guides(size = FALSE) (@erdeyl).tidyr::gather() with tidyr::pivot_longer() in ggballoonplot() and compare_means() internals (#536, @erdeyl).dplyr::do() with dplyr::reframe() in compare_means(), desc_statby(), and internal helpers. Replaced dplyr::mutate_if() with dplyr::mutate(across(where())) in ggsummarytable() (@erdeyl).border() deprecation warning by using linewidth instead of size in element_rect() (#644, #654, #656, @erdeyl).size deprecation warnings in ggscatter() (rug and star plots), ggpaired() (connecting lines), ggecdf() (ECDF line), ggdensity() (density lines), geom_bracket(), and geom_pwc() (#645, @erdeyl).stat_cor() parsing error when options(OutDec = ",") is set (European decimal separator) by using decimal.mark = "." in formatC() calls (#512, @erdeyl).compare_means() error "object 'group2' not found" when using ref.group with method = "anova" or method = "kruskal.test" (#572, @erdeyl).exact = FALSE workaround from version 0.6.2 that forced non-default behavior on wilcox.test(). Tests now use flexible assertions to ensure compatibility across R versions (#649, #647)..parse_font() not recognizing decimal font sizes (e.g., lab.font = c(2.4, "italic", "black")), which caused label colors to render incorrectly (#659).compare_means() function now sets exact = FALSE for wilcox.test() and pairwise.wilcox.test() to maintain backward compatibility and consistent p-values across R versions (#647).outliers parameter to ggboxplot() to control the display of outlier points. Set outliers = FALSE to remove the black dots representing outliers from box plots (#614, @hswl1314)...p.signif.., ..eq.label..) to modern after_stat() calls with proper namespace qualification.ggline() parameter handling for ggplot2 3.4.0+ compatibility:
linewidth parameter for line widthsize parameter for lines with helpful warning messagesize parameterafter_stat() calls that were causing failures in reverse dependency packages (bSi, PopComm). The convert_label_dotdot_notation_to_after_stat() function now properly handles namespace qualification while maintaining backward compatibility (#638).ggplot2::after_stat() is accessible during plot building, resolving "could not find function after_stat" errors in downstream packages.stat_regline_equation() to display in standard mathematical convention "y = mx + b" instead of "y = b + mx" (#559, @tshates, @mwaak).gghistogram() tests to handle changes in binning standardization introduced in ggplot2 4.0.0 (#635, @teunbrand).stat_pvalue_manual() failing when fill or other aesthetics are provided in the parent ggplot layer. The function now sets inherit.aes = FALSE by default to prevent conflicts between parent plot aesthetics and the p-value annotation data (#621, @fncokg).ggplot2::is.ggplot() with ggplot2::is_ggplot() in ggpar().data$column syntax to quoted column names in geom_pwc() for tidyselect 1.2.0+ compatibilityall_of() wrapper in unnest() utility function for tidyselect compatibilitysize by linewidth in ggplot2 element_line() and element_rect() functions.stat_regline_equation() by automatically converting deprecated dot-dot notation (..eq.label.., ..adj.rr.label.., ..p.signif.., etc.) to after_stat() syntax for ggplot2 3.4.0+ compatibility (#623, @hinkyisme).add_summary() and ggerrorplot() for ggplot2 compatibility:
ggadjust_pvalue() added to adjust p-values produced by geom_pwc() on a ggplot (#522).gene_expressionggpubr.null_device, whose value should be a function that creates an appropriate null device. These include: cowplot::pdf_null_device, cowplot::png_null_device, cowplot::cairo_null_device and cowplot::agg_null_device. Default is cowplot::pdf_null_device. This is used in functions like as_ggplot(), which need to open a graphics device to render ggplot objects into grid graphics objects. This function is used to open a null device to avoid displaying an unnecessary blank page when calling ggarrange() or as_ggplot() (#306 and #158). The default option can be changed using, for example, options(ggpubr.null_device = cowplot::png_null_device).gadd(): Restoring back random state after setting seed when adding jittered points. To do so, the seed number is just passed to position_jitter() and position_jitterdodge(), which preserve the initial random state ( #177 and #349) .ggpubr now requires a version of ggrepel >= 0.9.2.9999, which now restores the initial random state after set.seed(). See https://github.com/slowkow/ggrepel/issues/228ggpubr now requires a version of cowplot >= 1.1.1ggtexttable(): doc updated with another example; text justification for individual cells/rows/columns (#335).ggpie(): setting the default of clip = "off" in coord_polar() so that ggpie() does not crop labels (#429)as_ggplot(): using null_device to avoid blank page #306 and #158ggarrange(): using null_device to avoid blank page #306 and #158df is a tibble.ggexport(): support added for graphics device svg (#469)ggpie() and ggdonutchart() now fully reacts to the option lab.font (#502)gather_() in both internal (.check_data()) and exported functions (compare_means()) (#513)stat_compare_means(): The dot-dot notation (..p.signif..) was deprecated in ggplot2 3.4.0; after_stat(p.signif) should be used; updated so that ..p.signif.. is automatically converted into after_stat() format without warning for backward compatibility.desc_statby() doc updated to clarify the difference between SD (standard deviation) and SE (standard error) (#492)geom_smooth() using formula 'y ~ x' is now turned off in ggscatter()(#488)ggtext(): fix warning "filter_() was deprecated in dplyr 0.7.0".ggqqplot(): the argument conf.int is taken into account now when specified (#524).ggqqplot(): Fixing the warning: "The following aesthetics were dropped during statistical transformation: sample" (#523)rstatix v >=0.7.1.999 for preserving factor class in emmeans_test() (#386)ggmaplot(): Suppressing ggmaplot warning: Unlabeled data points (too many overlaps). Consider increasing max.overlaps (#520)compare_means(): works now when the grouping variable levels contain the key words group2 or group1 (#450)ggparagraph() : fixing bug about minimum paragraph length (#408)ggexport(): the verbose argument is now considered when specified by user (#474)stat_anova_test(), stat_kruskal_test(), stat_welch_anova_test(), stat_friedman_test() and geom_pwc() added. These are flexible functions to add p-values onto ggplot with more options. The function geom_pwc() is for adding pairwise comparisons p-values to a ggplot; supported statistical methods include "wilcox_test", "t_test", "sign_test", "dunn_test", "emmeans_test", "tukey_hsd" and "games_howell_test".as_npc(), npc_to_data_coordinates() and get_coord().ggpubr_options() to display allowed global options in ggpubrggpubr.parse_aes. Logical indicating whether to parse aesthetic variable names. Default is TRUE. For example, if you want ggpubr to handle non-standard column names, like A-A, without parsing, then set this option to FALSE using options(ggpubr.parse_aes = FALSE).stat_conf_ellipse: ensure stat returns a data.frame for compatibility with ggplot2 v>=3.4.0stat_compare_means():
after_stat(p.signif) as the dot-dot notation (..p.signif..) was deprecated in ggplot2 3.4.0 (#509).ggdensity() and gghistogram(): dot-dot notation (..density.., ..count..) replaced by after_stat(density) and after_stat(count), respectively for compatibility with ggplot2 3.4.0.create_aes():
digits and table.font.size) added to ggsummarystats() for changing the summary table decimal place and text size (#341).stat_pvalue_manual() the argument hide.ns can be either a logical value (TRUE or FALSE) or a character value ("p" or "p.adj" for filtering out non significant by p-value or adjusted p-values).vjust = 0.5 (#301).Capital NS. is no longer displayed by stat_compare_means() (#171)ggshistogram() to make sure that it works when:
after_stat(),ggscatter() to make sure that:
ggpubr.parse_aes global option is set to FALSE (#229)ggtexttable() (#125, #129 and #283):
tab_cell_crossout(): cross out a table cell.tab_ncol(), tab_nrow(): returns, respectively, the number of columns and rows in a ggtexttable.tab_add_hline(): Creates horizontal lines or separators at the top or the bottom side of a given specified row.tab_add_vline(): Creates vertical lines or separators at the right or the left side of a given specified column.tab_add_border(), tbody_add_border(), thead_add_border(): Add borders to table; tbody is for table body and thead is for table head.tab_add_title() and tab_add_footnote() to add titles and footnotes (#243).create_aes() added to create aes mapping from a list. Makes programming easy with ggplot2 (#229).coord.flip added to support adding p-values onto horizontal ggplots (#179). When adding the p-values to a horizontal ggplot (generated using coord_flip()), you need to specify the option coord.flip = TRUE.median_hilow_() and median_q1q3() - added (@davidlorenz, #209):
median_hilow_(): computes the sample median and a selected pair of outer quantiles having equal tail areas. This function is a reformatted version of Hmisc::smedian.hilow(). The confidence limits are computed as follows: lower.limits = (1-ci)/2 percentiles; upper.limits = (1+ci)/2 percentiles. By default (ci = 0.95), the 2.5th and the 97.5th percentiles are used as the lower and the upper confidence limits, respectively. If you want to use the 25th and the 75th percentiles as the confidence limits, then specify ci = 0.5 or use the function median_q1q3().median_q1q3(): computes the sample median and, the 25th and 75th percentiles. Wrapper around the function median_hilow_() using ci = 0.5.get_breaks() added to easily create breaks for numeric axes. Can be used to increase the number of x and y ticks by specifying the option n. It's also possible to control axis breaks by specifying a step between ticks. For example, if by = 5, a tick mark is shown on every 5 (@Chitanda-Satou, #258).ggscatterhist() (@juliechevalier, #176):
ggscatterhist() is now a list of ggplots, containing the main scatter plot (sp) and the marginal plots (xplot and yplot), which can be customized by the end user using the standard ggplot verbsalternative supported in stat_cor() (#276).position in ggline() to make position "dodged" (#52).outlier.shape in ggboxplot(). Default is 19. To hide outlier, specify outlier.shape = NA. When jitter is added, then outliers will be automatically hidden.ggdotchart() using the option sorting = "none" (#115, #223).weight added in gghistogram() for creating a weighted histogram (#215)ggscaterhist() takes into account the argument position in margin.params when marginal plot is a histogram (#286).ggbarplot() enhanced to better handle the creation of dodged bar plots combined with jitter points (@aherholt, #176)bracket.shorten added in stat_pvalue_manual() and geom_bracket(). a small numeric value in [0-1] for shortening the with of bracket (#285).bracket.nudge.y added in stat_pvalue_manual() and geom_bracket(). Vertical adjustment to nudge brackets by. Useful to move up or move down the bracket. If positive value, brackets will be moved up; if negative value, brackets are moved down (#281).numeric.x.axis added in ggerrorplot(); logical value, If TRUE, x axis will be treated as numeric. Default is FALSE (#280).width is now considered in ggadd() for plotting error bars (#278).linetype in ggpaired().geom_exec() used in ggpaired() to add lines between paired points.ggmaplot() now supports two input formats (#198):
ggmaplot():
alpha for controlling point transparency/density (@apcamargo, #152).label.select to select specific genes to show on the plot (@apastore, #70)ggadd() the fill argument is considered for jitter points only when the point shape is in 21:25 (@atakanekiz, #148).parse added in ggscatter() and in ggtext(). If TRUE, the labels will be parsed into expressions and displayed as described in ?plotmath (#250).stroke supported in ggscatter() and in ggline(). Used only for shapes 21-24 to control the thickness of points border (@bioguy2018, #258).stat_cor() function code has been simplified. New arguments p.accuracy and r.accuracy added; a real value specifying the number of decimal places of precision for the p-value and the correlation coefficient, respectively. Default is NULL. Use (e.g.) 0.01 to show 2 decimal places of precision (@garthtarr, #186, @raedevan6, #114, #270).annotate_figure() manual updated to show how to use of superscript/subscript in the axis labels (#165).ggtextable() now supports further customization when theme is specified (#283).font.family is now correctly handled by ggscatter() (#149)ggpar() arguments are correctly applied using ggpie() (#277).ggscatter(): When conf.int = FALSE, fill color is set to "lightgray" for the regression line confidence band (@zhan6073, #111).gghistogram() supports the parameter yticks.by (@Chitanda-Satou, #258).ggsummarystats() to create a GGPlot with summary stats table under the plot ( #251).clean_table_theme() to clean the theme of a table, such as those created by ggsummarytable()ggbarplot() now supports stacked barplots with error bars (#245).vjsut in stat_compare_means() to move the text up or down relative to the bracket.type in geom_bracket() to specify label type. Can be "text" or "expression" (for parsing plotmath expression); #253.labeller to the function facet()position in get_legend() to specify legend positionlegend.grob in ggarrange() to specify a common legend you want to add onto the combined plot.cor.coef.name in the function stat_cor(). Can be one of "R" (pearson coef), "rho" (spearman coef) and "tau" (kendall coef). Uppercase and lowercase are allowed (@andhamel, #216).digits, r.digits, p.digits in the function stat_cor(). Integer indicating the number of decimal places (round) or significant digits (signif) to be used for the correlation coefficient and the p-value (@raedevan6, #216).compare_means() adapted to tidyr v>= 1.0.0 by specifying cols in the unnest() function (@Youguang, #216).stat_pvalue_manual() can now handle an rstatix test result containing only one group column.stat_central_tendency() to add central tendency measures (mean, median, mode) to density and histogram plotsstat_overlay_normal_density() to overlay normal density plot (with the same mean and SD) to the density distribution of 'x'.exact = FALSE is no longer used when computing correlation in stat_cor() (@tiagochst, #205)ggpie() keeps now the default order of labels (@WortJohn, #203)geom_bracket() for adding brackets with label annotation to a ggplot. Helpers for adding p-value or significance levels to a plot.compare_means() has been adapted to tidyr v1.0.0 (@jennybc, #196)geom_exec() now handles geom_bracket() argumentsvjust, hide.ns, step.increase, step.group.by, color and linetype added in stat_pvalue_manual()stat_pvalue_manual() can now guess automatically the significance label column.show.legend added to ggadd() and add_summary() functions.gghistogram(). Works now when the x variable is R keyword, such as var, mean, etc. (#192)ggline(), error bars now react automatically to grouping by line type (#191)step.increase added in stat_compare_means() to avoid overlap between brackets.stat_pvalue_manual() x axis variable is no longer automatically converted into factor. If your x variable is a factor, make sure that it is converted into factor.stat_pvalue_manual() can automatically handle the output of rstatix testsggbarplot() and ggviolin() now automatically create error bars by groups when users forget the option add.params = list(group = ) (#183).ggarrange() works when either ncol = 1 or nrow = 1 (@GegznaV, #141.compare_means() set automatically the option exact = FALSE. This is no longer the case (@stemicha, #141.stat_pvalue_manual() now supports dodged grouped plots (@emcnerny, #104).position is now handled by ggdotplot() (@Adam-JJJJJ, #178)label.sep argument works now in ggscatter() and stat_cor() (@sbbmu, #150)ggscatter() to avoid freezing when the add argument is incorrect (@atakanekiz, #135).The option ref.group was only considered when the grouping variable contains more than two levels. In that case, each level is compared against the specified reference group. Now, ref.group option is also considered in two samples mean comparisons (@OwenDonohoe, #118)
Now, ggqqplot() reacts to the argument conf.int.level (@vsluydts, #123)
Added error bar color is now inherited from the main plot (@JesseRop, #109)
bxp.errorbar added to ggboxplot() for adding error bars at the top of the box plots (@j3ypi, #105.stat_pvalue_manual() for adding p-values generated elsewhere (@achamess, #81, @grst, #65).alpha option added to ggviolin() @mtmatter, #77bracket.size added to stat_compare_means() @mtmatter, #43stat_cor() supports R^2 as an option @philament, #32position added in gghistogram(). Allowed values include "identity", "stack", "dodge".ci added in ggerrorplot() @abrar-alshaer, #94ggscatter() can remove the letter 'a' from the legend, when the argument show.legend.text = FALSE specified @atsyplenkov, #106.size option to ggscatter add.params is supported @retrogenomics, #94.ggdonutchart() added.Significance levels can be now customized and passed to stat_compare_means() (@jaison75, #45).
Editing pdf size is now supported in ggexport() (@JauntyJJS, #45).
ggscatterhist() the x variable was plotted two times, on both the plot x & y margins, instead of having, as expected, a) the x variable on the main plot x margin and 2) the y variable on the main plot y margin. This has been now fixed.ggdotchart() sorted automatically within groups when the color argument is specified, even when groups = NULL. This default behaviour has been now removed. Sorting within groups is performed only when the argument group is specified (@sfeds, #90).yticks.by and xticks.by work with NAs (@j3ypi, #89).New function ggballoonplot() added to visualize a contingency table.
ggdotchart() can be now used to plot multiple groups with position = position_dodge() (@ManuelSpinola, #45).
New function ggscatterhist() to create a scatter plot with marginal histograms, density plots and box plots.
New theme theme_pubclean(): a clean theme without axis lines, to direct more attention to the data.
New arguments in ggarrange() to customize plot labels (@G-Thomson, #41):
New argument method.args added to stat_compare_means(). A list of additional arguments used for the test method. For example one might use method.args = list(alternative = "greater") for wilcoxon test (@Nicktz, #41).
New argument symnum.args added to stat_compare_means(). A list of arguments to pass to the function symnum for symbolic number coding of p-values. For example, symnum.args <- list(cutpoints = c(0, 0.0001, 0.001, 0.01, 0.05, 1), symbols = c("****", "***", "**", "*", "ns"))
New functions table_cell_font() and table_cell_bg() to easily access and change the text font and the background of ggtexttable() cells (@ProbleMaker, #29).
New argument numeric.x.axis in ggline(). logical. If TRUE, x axis will be treated as numeric. Default is FALSE. (@mdphan, #35)
New argument lab.nb.digits in ggbarplot(). Integer indicating the number of decimal places (round) to be used (#28). Example: lab.nb.digits = 2.
New argument tip.length in stat_compare_means(). Numeric vector with the fraction of total height that the bar goes down to indicate the precise column. Default is 0.03. Can be of same length as the number of comparisons to adjust specifically the tip length of each comparison. For example tip.length = c(0.01, 0.03).
get_legend() returns NULL when the plot doesn't have legend.Now data argument are supported in stat_compare_means() when the option comparisons are specified (@emcnerny, #48)
Now compare_means() returns the same p-values as stat_compare_means() (@wydty, #15).
stat_compare_means() now reacts to label = "p.format" when comparisons specified (#28).
Now, the p.values are displayed correctly when ref.group is not the first group (@sehufnkjesktgna, #15).
In ggpar(), now legend.title can be either a character vector, e.g.: legend.title = "Species" or a list, legend.title = list(color = "Species", linetype = "Species", shape = "Species").
New argument ellipse.border.remove in ggscatter() to remove ellipse border lines.
ggscatter(mtcars, x = "mpg", y = "wt",
color = "cyl",
ellipse = TRUE, mean.point = TRUE,
ellipse.border.remove = TRUE)
In ggscatter(), the argument mean.point now reacts to fill color.
Support for text justification added in ggtexttable() (@cj-wilson, #15)
The function ggpie() can now display japanese texts. New argument font.family in ggpie() and in ggpar() (@tomochan001, #15).
Using time on x axis works know with ggline() and ggbarplot() (@jcpsantiago, #15).
stat_compare_means() now reacts to hide.ns properly.drawDetails.splitText() exported so that the function ggparagraph() works properly.ggbarplot(), now labels correspond to the true size of bars (@tdelhomme, #15).stat_compare_means() now keeps the default order of factor levels (@RoKant, #12).New helper functions:
gradient_color() and gradient_fill(): change gradient color and fill palettes.clean_theme(): remove axis lines, ticks, texts and titles.get_legend(): to extract the legend labels from a ggplot object.as_ggplot(): Transform the output of gridExtra::arrangeGrob() and gridExtra::grid.arrange() to an object of class ggplot.ggtexttable(): to draw a textual table.ggparagraph(): to draw a paragraph of text.annotate_figure() to annotate (arranged) ggplots.text_grob() to create easily a customized text graphical object.background_image() to add a background image to a ggplot.New theme function theme_transparent() to create a ggplot with transparent background.
gghistogram(), density curve and rug react to the fill color.ggarrange():
align to specify whether graphs in the grid should be horizontally ("h") or vertically ("v") aligned.legend to remove or specify the legend position when arranging multiple plots.common.legend to create a common unique legend for multiple plots.New functions:
ggarrange() to arrange multiple ggplots on the same page.ggexport() to export one or multiple ggplots to a file (pdf, eps, png, jpeg).ggpaired() to plot paired data.compare_means() to compare the means of two or multiple groups. Returns a data frame.stat_compare_means() to add p-values and significance levels to plots.stat_cor() to add correlation coefficients with p-values to a scatter plot.stat_stars() to add stars to a scatter plot.Now, the argument y can be a character vector of multiple variables to plot at once. This might be useful in genomic fields to plot the gene expression levels of multiple genes at once. see ggboxplot(), ggdotplot(), ggstripchart(), ggviolin(), ggbarplot() and ggline.
The argument x can be a vector of multiple variables in gghistogram(), ggdensity(), ggecdf() and ggqqplot().
New functions to edit ggplot graphical parameters:
font() to change the appearance of titles and labels.rotate_x_text() and rotate_y_text() to rotate x and y axis texts.rotate() to rotate a ggplot for creating horizontal plot.set_palette() or change_palette() to change a ggplot color palette.border() to add/change border lines around a ggplot.bgcolor() to change ggplot panel background color.rremove() to remove a specific component from a ggplot.grids() to add grid lines.xscale() and yscale() to change axis scale.New helper functions:
facet() added to create multi-panel plots (#5).add_summary() to add summary statistics.ggadd() to add summary statistics or a geometry onto a ggplot.New data set added: gene_citation
New arguments in ggpar(): x.text.angle and y.text.angle
New arguments in ggpubr functions, see ggboxplot(), ggdotplot(), ggstripchart(), ggviolin(), ggbarplot() and ggline:
combine added to combine multiple y variables on the same graph.merge to merge multiple y variables in the same plotting area.select to select which item to display.remove to remove a specific item from a plot.order to order plot items.label, font.label, label.select, repel, label.rectangle to add and customize labelsfacet.by, panel.labs and short.panel.labs: support for faceting and customization of plot panelsNew argument grouping.vars in ggtext(). Grouping variables to sort the data by, when the user wants to display the top n up/down labels.
New arguments in theme_pubr():
palette can be also a numeric vector of length(groups); in this case a basic color palette is created using the function grDevices::palette().Now, ggpar() reacts to palette when length(palette) = 1 and palette is a color name #3.
ggmaplot() now handles situations, where there is only upregulated, or downregulated genes.
get_palette() to generate a palette of k colors from ggsci palettes, RColorBrewer palettes and custom color palettes. Useful to extend RColorBrewer and ggsci to support more colors.ggpar() function can handle a list of ggplots.right.show.legend.text in the ggscatter() function. Use show.legend.text = FALSE to hide text in the legend.title, submain, subtitle, caption, font.submain, font.subtitle, font.caption in the ggpar() function.font.family in ggscatter().ggdensity (gghistogram) are now shown when data have NA values @chunkaowang, #1ggtext() for textual annotation.ggscatter(). A logical value. If TRUE, a star plot is generated.geom_exec(). A helper function used by ggpubr functions to execute any geom_xx functions in ggplot2. Useful only when you want to call a geom_xx function without worrying about the arguments to put in ggplot2::aes().ggbarplot().
theme_classic2() added. Classic theme with axis lines.ggboxplot(), ggviolin(), ggdotplot(), ggstripchart(), gghistogram(), ggdensity(), ggecdf() and ggqqplot() can now handle one single numeric vector.# Example
ggboxplot(iris$Sepal.Length)
gghistogram(), when add_density = TRUE, y scale remains = "..count..".