Updates to account for changes in ggplot2 v4.0.0. (#241)
Improved messaging when users pass competing risks models to
add_pvalue() when more than one outcome is on
display.
Added an example to the gallery for combining multiple survival endpoints. (#212)
Fixed confidence interval labels being incorrectly swapped for
multi-state models in tidy_survfit(). Multi-state models
now preserve the correct ordering of conf.low and
conf.high from the underlying broom::tidy()
output. (#215)
Improved error message when ggcuminc() is called
with survfit.coxphms objects. (#222)
Fixed add_risktable() to respect x-axis breaks when
x-axis is duplicated. (#221)
Correcting an argument name partial match in
ggplot2::scale_*(labels). (#211, @DanChaltiel)
We now allow for negative follow-up times in
tidy_survfit() (and subsequently ggsurvfit()).
When negative follow-up times are present users should specify
survfit(start.time) and we print a note to this effect when
not set. (#192)
The tidy_survfit() (and subsequently
ggsurvfit()) now honor the survfit(start.time)
if specified. (#192)
Updated legend position syntax to account for changes in {ggplot2} v3.5.0.
As of {survival} v3.6-4, the number censored are now returned as
a matrix for multi-state models (i.e. competing risks models). The
tidy_survfit() function has been updated to account for
this new structure and the minimum version of {survival} has been
increased to the latest version. (#199)
For non-competing risks multi-state models created with
survfit(), the y-axis label is now “Probability in State”.
(#205)
Added the “cloglog” transformation option to
ggsurvfit(type) and tidy_survfit(type).
(#194)
By default, a model plot created with ggsurvfit() or
ggcuminc() uses the color aesthetic to plot curves by the
stratifying variable(s), and further, ggcuminc() uses the
linetype aesthetic for plots that contain multiple outcomes
(i.e. competing events). We now introduce the global option
"ggsurvfit.switch-color-linetype" to switch these defaults,
giving users more flexibility over the output figures. Furthermore, when
the linetype_aes= argument is called in a situation when it
does not apply, it will be silently ignored (previously, an error
message may have been thrown). (#166)
Slightly increased the padding to the right of the plot when
scale_ggsurvfit() is called. (#165)
Now exporting the utility function
ggsurvfit_align_plots() that is used to ensure risktable
and the primary plot align. By exporting this function, users will now
be able to construct custom risktable graphics. See
?ggsurvfit_align_plots() for an example. (#175)
tidy_survfit() that change the
monotonicity of the curve, the conf.low and
conf.high column names are now switched. (#154)Added feature in add_risktable(risktable_stats=) to
accept glue-like syntax—anything inside curly brackets will be
evaluated. Users can now place multiple types of statistics on the same
row, including allowing users to style the statistics in any way they
like by adding rounding/formatting functions within the curly brackets.
Users may now also display estimates and confidence limits in the risk
table. (#135)
Updated ggsurvfit(), tidy_survfit(),
and survfit2() to handle survival::coxph()
models. (#9)
Updated the default margin size when a risktable is added to a figure. Namely, the primary plot’s bottom margin is set to zero, along with each risktable’s top and bottom margin. Moreover, the top and bottom margin of the legend is also set to zero.
All calls to aes() have been migrated from
ggplot() to the individual geoms,
e.g. geom_step(). This was done because adding the
aes() call in ggplot() led to an error when a
later geom is added with user-created data. (#127)
Delay adding the conf.low and conf.high
to the ggplot aes() until
add_confidence_interval() has been called. Previously,
these were being added in the first call to
ggplot(tidy_data, aes(x = time, y = estimate, ymin = conf.low, ymax = conf.high)).
The result was that figures that did not show the confidence
interval still created space for the CI in the plot area. This update
eliminates that blank space. (#123)
Migrated the ‘scales’ package from ‘Imports:’ to ‘Suggests:’, i.e. from a strong to a weak dependency. (#120)
Updated ggplot size= argument to
linewidth= where needed as of ggplot2 v3.4.0
(#131)
ggsurvfit_build(combine_plots=) argument to
TRUE.Added function add_pvalue() to place p-values in the
figure caption or as a text annotation.
Added the add_quantile(x_value=) argument that
places line segments at the time specified.
Added the scale_ggsurvfit() function that wraps both
ggplot2::scale_x_continuous() and
ggplot2::scale_y_continuous() and uses reduced padding (via
the expand= argument), labels y-axis with percentages
(labels=), adds additional break points on the x-axis
(n.breaks=8), and sets the y-axis limits to
c(0, 1) (limits=). (#82)
Added function add_legend_title() that adds a title
for the strata in the figure legend.
Package now depends on {ggplot2}, meaning that it’ll be attached anytime {ggsurvfit} is attached. (#62)
Added S3 methods grid.draw.ggsurvfit() and
grid.draw.ggcuminc() which in turn allows us to save images
from the package directly with ggplot2::ggsave()
(#107)
Added support for multi-state models created with
survfit(), i.e. competing risks from the survival package.
(#83)
Updated the default behavior of
add_risktable(risktable_group='auto') to minimize the
number of risk tables that appear below the figure. (#117)
Increased the default font size on the plot and in the risk tables, and added arguments to control font size in the risk table theme. (#103)
When using a CDISC ADTTE data frame, the label saved in
PARAM/PARAMCD will be used as the default x-axis label in
ggsurvfit(). (#97)
When the survfit(weights=) argument is utilized, the
number at risk, number of observed events, etc. are a non-integer
numbers. The counts in the risk table are now rounded to the nearest
integer. (#90)
Converted the gallery vignette to an article. (#75)
Bug fix when Surv_CNSR() is used in conjunction with
ggsurvfit(). The default x-axis label is incorrectly
attributed to a stratifying variable, when present. (#100)
Fix in survfit2() that allows users to pass
arguments with non-standard evaluation, i.e. bare column names.
(#90)