Full reference to all functions available at grafify
GitHub pages.
This version has major updates to linear mixed effects models
functions (mixed_model, mixed_anova,
mixed_model_slopes, mixed_anova_slopes). The
other major updates are to plot_density and
plot_histogram, which can now plot probability density,
counts or normalised counts (only latter two options for histograms).
Most old code should still work, but care is needed if arguments are not
explicitly called. Changes as follows:
Updates to mixed_model, mixed_anova,
mixed_model_slopes, mixed_anova_slopes:
AvgRF, which is
set to TRUE by default. This means that values of the response variable
(Y_value) are averaged before a linear mixed effects model
is fitted. This avoids pseudoreplication if there are multiple
‘technical replicates’ within levels of the Random_Factor.
For behaviour like before (v4.0.1 or lower), set
AvgRF = FALSE.Formula argument allows the user to directly
provide a formula to lmer(), which may be useful in more complex
scenarios (e.g., nested designs). This argument is available in
mixed_model and mixed_anova functions.Y_value or
Fixed_Factor: log(x), log(x + c), logit(x), log(x/100),
sqrt() or exponents (e.g., (c)^2).Updates to arguments in plot_density and
plot_histogram:
PlotType to choose a
type of data transformation, i.e., use ggplot2 defaults or
plot counts or normalised counts. For plot_density, this
argument can be Density (default), which plots the smooth
probability density curve, Counts or
Normalised counts. For plot_histogram, the
default for this argument is Counts, which can be changed
to Normalised counts. This is achieved by passing
y = after_stat(count) or
y = after_stat(count/max(count)) to
geom_density or geom_histogram.plot_ functions in grafify, these arguments
are LogYTrans, LogYBreaks,
LogYLabels and LogYLimits.SingleColour argument has been added to change
colours when a grouping variable is missing, i.e., when the argument
group is not provided with a value. If a value is provided,
however, SingleColour is ignored and the
ColPal value is applied.plot_.. functions, the
LogYLabels argument was called Ylabels, which
has now been deprecated for consistency (old code will still work).ggplot2 v3.5, the discrete_scale()
has been updated and an argument (scale_name) has been
dropped. The scale_fill_grafify and
scale_colour_grafify functions have been accordingly
updated.geom_violin (i.e.,
plot_scatterviolin, plot_3d_scatterviolin and
plot_4d_scatterviolin), drop = FALSE has been
added to avoid throwing up errors. Be careful if you have fewer than 2
data points.This is a minor update to fix tests that were failing after an update
to ggplot2 to version 3.5.0.
The reason this is a major update is that now ggplot2
version 3.4.0 or higher is required to avoid errors with
grafify. The main difference is that size
argument for line widths has been updated to linewidth.
It is easier to plot 2-way ANOVA designs with or without blocking
factors in this version with the following updates. There are two new
plot_... functions for 1-way and 2-way designs.
plot_4d_ functions can now plot 2-way ANOVAs even if
the shapes argument is not provided. Graph is plotted with
shape = 21 as default.
plot_4d_point_sd and plot_3d_point_sd
functions for plotting 2-way and 1-way ANOVAs without or with blocking
factors as mean and SD/SEM/CI95 error bars.
{{ in theme_grafify.hjust and vjust arguments to
theme_grafify to adjust text alignment when angles are
changed.ggplot2
update.theme_grafify updates:
lineend = square as default for better-looking
origingrafify plots by
defaultbasesize (default 20)New arguments in violin plots (plot_dotviolin,
plot_scatterviolin, plot_3d_scatterviolin and
plot_4d_scatterviolin): two separate arguments
bthick and vthick to set the line widths of
the boxes and violins, respectively. The previous bvthick
will still work, so if a value is provided that will be used for line
widths of both boxes and violins.
New argument for two-way ANOVA graphs (plot_4d_):
the group_wid can be used to change the space between
groups along the X-axis (i.e., dodge width). Default
group_wid = 0.8 will produce graphs that look similar to
those in previous versions of grafify. If
group_wid is set to 0, there will be no dodging of the
factors along X-axis.
New arguments in before-after plots (i.e.,
plot_befafter_): bthick and
lthick arguments can change line and box line widths
independently.
bwid argument) of
bars and boxes in plot_4d_ functions is set as
0.7.log10 tick marks:
plot_ functions: the tick marks now scale with
the fontsize parameter. Previously, the sizes were set to
“cm” units, which did not scale correctly. The long tick mark, middle
and short ticks are sized: 7*fontsize/22,
4*fontsize/22 and 4*fontsize/22, respectively
(note that the short and mid are the same size). The size (line width)
equals fontsize/22, which is the same throughout
grafify.plot_logscales
function also has the above defaults and now has
fontsize = 20 as an additional argument and sizes scale
accordingly.log10 tick marks have the same colour as
ticks on non-transformed axis (grey20).plot_point_sd now allows all data points to be shown.
All points will be plotted with geom_point if the
all_alpha setting (opacity for all symbols) is set >0
(it is set to 0 so default graphs will only show the mean of all
values). There are also options for all_size and
all_jitter to adjust size and overlap.ErrorType argument (default is “SD” error bars) in
plot_dotbar_sd, plot_scatterbar_sd,
plot_point_sd, plot_3d_scatterbar and
plot_4d_scatterbar.plot_lm_predict which used to
label Y-axis as pred rather than the correct name of the
plotted variable.theme_grafify.plog_qqline based on stat_qq
and stat_qqline.plot_histogram which was throwing up
warning messages after ggplot2 update. For uniformity with
other grafify graphs, histograms now have a black border
(like symbol borders in dot/scatter plots).groups argument, which was deprecated several versions
before, has been removed from before-after functions.plot_bar_sd deprecated as similar graphs can be plotted
with plot_scatterbar_sd with s_alpha = 0.scale_colour_grafify, which broke in
v3.0.0.SingleColour argument can now take base R colour names
(e.g., “grey25”) in addition to previously available options.plot_grafify_palette can now also plot the quantitative
colour schemes.This is a major update for grafify, which now provides
wrappers for basic generalised additive models (gam)
through the mgcv
package. There are a more plot_ functions, a
grafify theme for ggplot objects, and simple
data wrangling before plotting. There are also updates within all
plot_ functions, which are a facet argument,
and log-transformations with axis tick marks.
Fit generalised additive models (gam) and get ANOVA tables with
two new functions: ga_model and ga_anova.
These are mainly for time-series analyses or where an assumption of
linear relationship between predictor and outcomes is absent straight
lines are not appropriate. Factor-wise smooths are fit with the
by argument in mgcv, without or with a random
factor. Random factors are also allowed with smooth re
smooth. See documentation for mgcv smooths. Model
diagnostics can be done with plot_qq_gam and
plot_qq_model. Example data included as
data_zooplankton is from Lathro
RC, 2000.
All plot_ functions now have two major updates:
log10 or log2 with LogYTrans and
LogXTrans arguments. X axis transformations are only
available for plot_xy_CatGroup and
plot_xy_NumGroup. With log10 transformation,
log-ticks will also appear. Default axes limits and labels should work
in most cases, but if needed, three additional arguments are available:
LogYBreaks, LogYLimits and
LogYLabels (and respective ones for the X axis).facet argument to add another variable to created
faceted plots with the facet_wrap layer in
ggplot2. A related argument facet_scales can
be used to set Y or X axis scales on faceted panels.New plot functions:
plot_befafter_box is a new before-after plot function
that includes a box and whisker plot to show data distribution in
addition to lines joining matched data. In addition, both
plot_befafter_colour and plot_befafter_shapes
offer a box and whiskers summary of data.plot_lm_predict and plot_gam_predict can
be used to plot observed (raw) data and predicted data from fitted
linear models.plot_logscales is a function to easily perform “log10”
or “log2” transformation on X or Y axes of any ggplot2
object along with log-ticks.Table manipulations:
table_x_reorder is a function to reorder levels within
a categorical variable. This uses factor from base R
stats package to convert a column into a factor and
reorders it based on a user-provided vector of group names.table_summary is a wrapper around
aggregate (base R) function, which gives mean, median, SD,
and counts grouped by one or more variables.A grafify theme for ggplot2:
theme_grafify is a modification of
theme_classic for making publication-ready
grafify-like graphs easily when using
ggplot2.
Much of the code has been edited and cleaned up. Among the main
change is dropping unnecessary double curly brackets {{ }}
within plot_ wrappers.
The main motivation behind this update was to simplify the package by
reducing the number of exported functions. So some features that were
previously in separate functions have been made available more easily
via an additional argument to existing functions (e.g. single colour
function (plot_..._sc) now offered in respective
plot_ function with a new argument (see below). This has
uncluttered the namespace of grafify. Most of the other
additions are related to colour schemes.
A new SingleColour argument has been added to
two-variables plot_ functions to generate graphs with a
single colour along the X-axis aesthetic. This means the 8
plot_..._sc functions introduced in v1.5.0 are deprecated,
but this feature is still retained in existing plot_
functions. This option also added to plot_3d_ functions for
plots of one-way ANOVA data.
Four new colourblind-friendly categorical colour schemes (chosen from cols4all package):
fishy, kelly, r4,
safeFour new quantitative schemes for continuous or divergent colours.
blue_conti,
grey_contiOrBl_div, PrGn_divAll schemes also available through scale_fill.. and
scale_colour_... calls to be used on any
ggplot2 object.
scale_fill_grafify and
scale_colour_grafify (or scale_color_grafify)
have been rewritten. These have two new arguments that offer features
previously in
scale_fill_grafify2/scale_colour_grafify2/
scale_color_grafify_c and
scale_fill_grafify_c/scale_colour_grafify_c/
scale_color_grafify_C scale functions. These 6 functions
are now deprecated to reduce exported namespace.The new arguments are discrete (logical T/F) to select
discrete or continuous palettes, and ColSeq (logical T/F)
to pick sequential or distant colours from a chosen palette.
plot_3d_ that incorrectly referred to xcol and
shapes arguments.plot_3d_scatterviolin as compared to the other two
plot_3d_ functions.posthoc_Trends... functions rewritten with
stats::model.frame() to get model data frame as this is a
more flexible method.light, bright and
muted schemes changed slightly for better separation of
colours when next to each other.jitter setting in plot_scaltter_ is
set to 0.2 so the graph as plotted with jitter by
default.okabe_ito (the all_grafify palette is was just
a concatenation of all palettes without real basis in good
visualisation). Use one of the other palettes if more than 8 colours are
needed (e.g. kelly, which has 20 discreet colours).plot_3d_scatterviolin and
plot_4d_scatterviolin for one-way or two-way ANOVA design
data to plot scatter plots with violins with box and whiskers.
plot_qqmodel no longer relies on
broom.mixed; instead uses rstudent from the
base stats package to generate studentized residuals from a
model.
New experimental functions to compare slopes of linear regression via
posthoc_Trends_Pairwise,
posthoc_Trends_Levelwise and
posthoc_Trends_vsRef.
Minor changes to plot_qqmodel and
plot_qqline to fix some OS-specific errors. QQ plots by
default will have ok_orange colour within symbols when only
one level is present within group. Both functions now use
geom_qq and geom_qq_line (instead of
stat_qq and stat_qq_line) internally.
This is a major update with some new features, bugfixes, and further
cleaning up of code with consistent names of arguments in preparation
for CRAN submission. Some previous code may not work because of renaming
of some arguments for grouping variables in plot_
functions. But older arguments are retained with deprecation warnings in
most cases, so old code should largely work.
plot_ functions have a new argument ColSeq
(logical TRUE/FALSE) that picks colours sequentially from palette chosen
by ColPal when TRUE (default). If set to
FALSE, the most distant colours are chosen, as already
implemented in scale_..._grafify2 functions.geom_violin. They also get
new arguments to set thickness of lines (bvthick) and
transparency of boxplots (b_alpha).mixed_model_slopes and
mixed_anova_slopes.posthoc_Trends implements the
emmeans::emtrends call.plot_ functions now have the ...
argument forwarding dots for advanced users to add arguments to
ggplot geometries where necessary.plot_grafify_palette function that helps quickly
visualise colours in palettes along with their names and hexcodes.plot_bar_sd and plot_bar_sd_sc have a new
argument bthick to adjust the thickness of lines of the
bars.Group grouping argument in
plot_density, plot_histogram and
plot_qqline is now called group for
consistency with other plot_ functions.Factor argument in post-hoc comparisons functions
(posthoc_Pairwise, posthoc_vsRes, and
posthoc_Levelwise) renamed as Fixed_Factor to
be consistent with mixed_model, simple_model,
mixed_anova and simple_anova functions.plot_3d_scatterbar and
plot_3d_scatterbox now correctly plot one-way ANOVA designs
with randomised blocks with shapes mapped to levels of the
random factor, and xcol as the grouping factor as
originally intended but incorrectly implemented. This complements
plot_4d_scatterbar and plot_4d_scatterbox
which take two grouping factors and a random factor.groups in before-after plots is now called
match as it is a bit more informative when showing matched
data.c_alpha in
plot_density and plot_histogram (for colour
opacity of colours under the density curve or histogram); opacity of
symbols in plot_qqline is still called
s_alpha.This update fixes and cleans up code to remove all errors, warnings
and notes from devtools::check(). All previous code should
still work.
broom.mixed::augment is used to
get model residuals than the fortify method as this will be
deprecated soon. The broom.mixed package therefore
required.lmerTest, but instead forces a mixed model
object as lmerModLmerTest object to get F and P values in
ANOVA tables from the stats::anova call.magrittr package is required for internal use of
pipes (%>%).simple_model and
mixed_model was cleaned up so that model outputs are as
close to objects generated by native calls to lm or
lmer.make_1w_rb_data and make_2w_rb_data
functions have been updated to have consistent factor and level
names.This version has 8 new plot_ functions ending in
_sc for plotting data with two variables wherein the X
variable is plotted in a single colour. This contrasts existing versions
that plot the X variable with multiple colours chosen from the
all_grafify palette. This is convenient when there are too
many groups on the X axis and multiple colours are not necessary.
plot_qqmodel will plot a diagnostic Q-Q plot of a simple
linear model (generated with simple_model or
lm) or mixed effects linear model (generated with
mixed_model or lmer) in a single step.
Fixed a typo in posthoc_Levelwise where the
adjust argument was not being correctly passed on to
emmeans.
This version “breaks” a few arguments from v0.3.1, therefore is
v1.4.1. Specifically, opacity for both symbols and bars/boxes/violins
can be set using s_alpha and b_alpha or
v_alpha, respectively; previously, only bars/boxes/violin
opacity could be set with a single alpha parameter. Old
code with just alpha will no longer work, sorry! There are
also new graph types and arguments for ANOVAs as below.
New graph types
plot_density and plot_histogram for smooth
density or histogram plots through geom_density and
geom_histogram respectively.plot_scatterbox and
plot_scatterviolin that complement the
plot_dot... versions and instead use
geom_point with position_jitter. These
versions are useful when a large number of data points are needed to be
plotted.Updates
simple_anova where the table also has Mean SS.mixed_anova now has two new arguments, one to change
method for Df calculation and second to get type I or III SS (default is
type II).jitter argument added to plot_3d.. and
plot_4d.. functions for consistency with other scatter
plots.bwid argument (for adjusting width of bars) added to
plot_scatterbar_sd for consistency.Bug fixes in mixed_model and simple_model
which now correctly lists the data used in the call field.
plot_4d_scatterbar function which is like
plot_4d_scatterbox but plots bar and SD. So there are now
two plot_3d_ and plot_4d_ functions.TextXAngle argument to prevent overlap.plot_dot_ functions now have dotthick
option to set stroke thickness. This is similar to symthick
for scatter/jitter plots.facet_wrap or facet_grid will not
draw a box around panel text (unlike the default in
theme_classic()).plot_3d_ and plot_4d_ functions draw
symbols in black colour.plot_3d_scatterbar and
plot_3d_scatterbox, which now correctly use the “shapes”
variable to fill colour of bars/boxes and shape of the symbols; symbols
are depicted in black.simple_anova generates type II ANOVA table through
car::Anova(), so the car package is now a
dependency. v0.1.0 and v0.2.0 generated type I ANOVA table through
stats::anova().plot_
functions apply the all_grafify colour scheme by default
(see plot_ vignettes on how to change colours)plot_xy_NumGroup) or categorical
(plot_xy_CatGroup).scale_fill_grafify_c and
scale_colour_grafify_c), based on Paul Tol’s
variant of YlOrBl scheme.plot_befafter... functions have a new logical
TRUE/FALSE argument called boxplot to additionally show a
box and whisker plot to show data distribution.First release.