The deprecated argument ppd was removed.
Some of the deprecated arguments in plot() were
removed.
Pooling functions now use the model’s degrees of freedom to calculate the critical values for the confidence intervals.
test = "slope" (or test = "trend") are
aliases in test_predictions() for test = NULL
with numeric predictors.
predict_response() (and ggpredict(),
ggemmeans() and ggeffect()) get an argument
bias_correction, to correct for bias when back-transforming
predictions for non-Gaussian mixed models.
glm_weightit,
ordinal_weightit, multinom_weightit from
package WeightIt.Fixed issues for multivariate response models due changes in the last insight package updates.
Fixed issue with swapped lower and higher confidence interval values for models with inverse-link.
Fixed CRAN check issues due to breaking changes in the last marginaleffects update.
ci.lvl in
test_predictions() was removed.ggpredict() now supports models of class
glmgee (package glmtoolbox).
ggemmeans() gains arguments vcov_fun,
vcov_type and vcov_args to specify the
variance-covariance matrix for the marginal means, similar to what is
already available in ggpredict() and
ggaverage().
When test = "contrast", the engine is
automatically set to "emmeans" in
test_predictions().
ggaverage() (or
predict_response(..., margin = "empirical")) now also
supports following type options for zero-inflated models:
"zi_prob", "zero_inflated" and
"fixed".
Support for zero-inflated models was massively improved in
test_predictions(), which, for instance, now also supports
scale = "zprob" to calculate contrasts for the
zero-inflation probability for zero-inflated models from
glmmTMB or pscl. Furthermore, when predictions for
zero-inflation probabilities were calculated using
pr <- predict_response(..., type = "zi_prob"),
corresponding contrasts will be calculated with
test_predictions(pr) automatically. Additionally, other
types for models with zero-inflation component
("zero_inflated", "fixed") are supported as
well.
ggeffect() now warns the user about arguments that
are supported by ggpredict() or ggemmeans(),
but not by this function (e.g., vcov_fun).
Improved accuracy of standard errors for
test_predictions(..., engine = "ggeffects").
The terms argument now also accepts the shortcut
"percentile" (plus numeric value) to select a range of
percentiles for continuous variables, e.g.
terms = "x [percentile90]" to select a range of the 90%
percentile.
Fixed issue with brms models with monotonic effects in
formula (mo()).
Fixed issue in vcov() for ggeffects
objects, which could occur in rare situations when some of the
predictors were character vectors.
Fixed issue with calculation of standard errors when one of the focal term was a character vector.
Fixed issue in plot() method with
show_data = TRUE, where in certain situations the raw data
points were not colored when groups were present.
Fixed issue in plot() with too many data points when
collapsing random effects groups.
ggpredict() now works for models of class
clm2 from package ordinal, however, confidence
intervals are not yet supported for these models.
ggeffect() now passes the latent
argument for models with ordinal outcome down to
effects::Effect(), to plot effects for ordinal models on
the latent scale.
When argument test in
test_predictions() is "interaction",
"consecutive", or a data frame, emmeans is
automatically used as backend, as this is the relevant package that
supports these argument types.
format() (and hence, print()) for
test_predictions() gains a combine_levels
argument, to combine levels of the focal term in the output
table.
The engine argument in
test_predictions() can now also be
"ggeffects". However, this is currently work-in-progress
and offers muss less options as the default engine,
"marginaleffects". It can be faster in some cases, though,
and works for comparing predicted random effects in mixed
models.
test_predictions() now automatically falls back to
engines "emmeans" or "ggeffects" if the
marginaleffects (or emmeans) package is not
installed.
predict_response(), test_predictions()
and ggpredict() will warn the user when all focal terms are
only included as random effects in the model and no appropriate
type or margin is specified. This is to avoid
meaningless results.
plot() gets an n_rows argument, to
define the number of rows for the panel alignment. This is useful when
the number of facets is large and the default alignment is not
optimal.
The ppd argument for Bayesian models will be
superseded by the interval argument,
i.e. ppd = TRUE is equivalent to
interval = "prediction" (and ppd = FALSE is
equivalent to interval = "confidence").
When back_transform = FALSE, and model has a
transformed response variable, the plot() method for
ggeffects objects now rescales the raw data points. This
ensures that the raw data points are plotted on the same scale as the
predicted values when show_data = TRUE.
Minor revisions of documentation and vignettes, to improve readability and clarity.
Several arguments have been deprecated and replaced by new argument names. A warning is printed when deprecated arguments are used. The deprecated arguments will be removed in a future release.
Fixed issue in print() for ggeffect()
and models with ordinal outcome, where one column was too much in the
output.
Fixed issue in test_predictions() with wrong order
of term labels when a focal term was a character vector.
Fixed issue in ggpredict() with wbm
models from package panelr.
Fixed issue in ggemmeans() for glmmTMB
models with zero-inflation, when terms included variables
that were specified in the conditional, but not in the zero-inflation
model formula.
Fixed issue in ggpredict() for Stan models (from
packages rstanarm and brms) where the
ci_level argument was not correctly recognized.
Fixed CRAN check issues due to latest marginaleffects update.
ggemmeans() (and hence,
predict_response(..., margin = "marginalmeans")) now
supports type = "zi_prob" for zero-inflated models from
package glmmTMB, i.e. can now predict the zero-inflation
probability.
test_predictions() and ggaverage() were
updated to work with the latest release of the marginaleffects
package. That release fixed issues with inaccurate standard errors for
glmmTMB models.
test_predictions() gains a margin
argument, to indicate how to marginalize over non-focal terms. This
ensures that estimates of pairwise comparisons are in line with
estimates of predictions.
test_predictions() gains an engine
argument, to indicate which package to use to compute pairwise
comparisons or contrasts. By default, the marginaleffects
package is used, but you can also use the emmeans
package.
ggeffect() when representative values
for a focal term included a zero,
e.g. terms = "focal [0,3,5]".Overhaul of the documentation (again), to provide more clarity about the terminology “adjusted predictions”, “marginal means” and “marginal effects”, and how to calculate each of these quantities using the ggeffects package.
print_html() methods were updated to work with the
latest release of tinytable.
New print_md() method, to print the output as
markdown table. This is useful inside RMarkdown or Quarto documents,
where the output can be directly included.
predict_response() as “generic” high-level function,
which is a replacement for ggpredict(),
ggemmeans() and ggaverage(). The new function
is more clear about how the function marginalizes over non-focal terms.
The margin argument can be used to specify how to
marginalize over non-focal terms, i.e. which function internally is used
to compute the marginal effects.The documentation was revised, to provide more clarity about what the package functions do and how to decide, which function or method to calculate marginal effects is the most appropriate.
Improved calculation of prediction intervals for Poisson regression models.
Improved handling of the vcov_fun argument. This
argument now accepts an estimation type as string,
e.g. vcov_fun = "HC0", which is then used to compute the
variance-covariance matrix. Thus, it is no longer necessary to define
both vcov_fun and vcov_type, if the
variance-covariance matrix is covered by one of the pre-defined
estimation types. See ?ggpredict for details.
hypothesis_test() now also accepts the
vcov_fun argument, and not only vcov. This
ensures consistency with the vcov_fun argument in
ggpredict(). Furthermore, the information about the type of
variance-covariance matrix is saved to the ggeffects object
returned by ggpredict(), predict_response()
etc., and if this information is available, it is automatically used in
hypothesis_test() when a ggeffects object is
passed to the function.
Fixed bug in wrong order of printed (sub-)tables for predictions.
Fixed wrong table column name for confidence interval columns for
other confidence levels than the default 95% in print() for
ggeffects objects.
Fixed issue with ggpredict() for models of class
fixest when the cluster variable was numeric.
print() method has been revised. A
format() method was added, which allows to format the
output of ggpredict() (and ggeffect() etc.)
for printing. The refactoring of the print() method makes
the code base easier to maintain and it is easier to enhance the
print-functionality. Now it is possible to create HTML tables as well,
using print_html(). The style of the output has also
slightly changed. By default, confidence intervals are no longer
enclosed in parentheses. You can change this behaviour by passing the
ci_brackets argument to print() (see
examples), or permanently define custom parentheses or brackets with,
e.g., options(ggeffects_ci_brackets = c("[", "]")).
Additionally, there are new arguments to further control the output of
the tables: collapse_ci can be used to collapse confidence
intervals into a single column together with the predicted values.
collapse_tables can be used to collapse multiple tables
into a single table (only applies when there is more than one focal
term). Again, these settings can be permanently defined via options (see
?print.ggeffects for details).print_html(), to print the output as HTML table. This
method is available for objects from ggpredict() (and
alike) as well as hypothesis_test().ggpredict() for models of class
vglm with multivariate responses.ggpredict() now supports models of class
rqs from package quantreg.
Fixed issues to be compatible with forthcoming update of emmeans.
ggaverage(), to compute average predicted values.
This function is based on
marginaleffects::avg_predictons().
pool_comparisons(), to pool results from multiple
calls to hypothesis_test(), e.g. with imputed data
sets.
Support for sdmTMB (sdmTMB)
models.
Improved support for the logistf package, including
models flic() and flac().
Confidence intervals for predictions from merMod
models (package lme4) now use the standard errors returned by
predict(..., se.fit = TRUE). This should not affect
numerical results, but can be more robust for certain edge
cases. Note that standard errors are only based on
predict() when tpye = "fixed". For
type = "random", standard errors are still based on the
model’s variance-covariance matrix, taking uncertainty from random
effects into account.
hypothesis_test() now suppports models from package
parsnip.
johnson_neyman() gains a p_adjust
argument, to adjust p-values for multiple comparisons. Currently, only
p_adjust = "esarey" (resp. p_adjust = "es")
and p_adjust = "fdr" (resp. p_adjust = "bh")
are supported.
ggpredict() now computes appropriate predicted
probabilites for models of class rms::lrm() with ordinal
outcome.
Fixed issue in ggpredict() for
type = "random" when sampling from random effects levels,
where the levels were numeric characters with a pattern like
"001", "002", etc.
Fixed minor issue in plot.ggalleffects().
... arguments in ggpredict() are now
passed down to the predict() method for
mgcv::gam() models.
The typical argument now supports a mix of functions
for different variable types at which numeric or categorical covariates
(non-focal terms) are held constant.
Clarification of how the re.form argument is set
when using type = "random" resp.
type = "fixed" in ggpredict().
hypothesis_test() now returns the standard error of
contrasts or pairwise comparisons as attribute
standard_error. This can be used to compute the
test-statistic, if required. In forthcoming updates, there will be
methods for insight::get_statistic() and
parameters::model_parameters() to include standard errors
and test-statistics in the output.
test_predictions() was added as an alias for
hypothesis_test().
hypothesis_test() for mixed models,
which sometimes failed when random effects group variables were numeric,
and not factors.johnson_neyman(), to create Johnson-Neyman intervals
and plots from ggeffects objects.Better automatic handling of offset-terms, both for predictions
and generating plots with raw data. When the model formula contains an
offset-term, and the offset term is fixed at a specific value, the
response variable is now automatically transformed back to the original
scale, and the offset-term is added to the predicted values. A warning
is printed when model contains transformed offset-terms that are not
fixed, e.g. via the condition argument.
ggeffect() now supports nestedLogit
models.
Fixed issue in hypothesis_test(), where the
by argument did not work together with the
collapse_levels argument.
Fixed issue in plot() method when adding raw data
points for data frame that had now row names.
jitter argument that is used to add some noice to data
points to avoid overlapping now defaults to NULL. Formerly,
a small jitter was added by default, leading to confusion when data
points did not match the original data.The plot() method gets a label.data
argument, to add row names to data points when
add.data = TRUE.
tibbles are always converted into data frames, to
avoid issues.
hypothesis_test() gains a by argument,
to specify a variable that is used to group the comparisons or
contrasts. This is useful for models with interaction terms.
Plotting residuals did not work when model object passed to
ggpredict() were inside a list, or when called from inside
functions (scoping issues).
Fixed issue where plotting raw data
(i.e. plot(..., add.data = TRUE)) did not work when there
were missing data in weight variables (i.e. when the regression model
used weights).
Fixes issue in plot() when no term was specified in
the call to ggpredict().
Fixed issues with robust estimation for models of package pscl.
Fixed issues introduced by breaking changes in marginaleffects.
Support for nestedLogit (nestedLogit)
models.
hyothesis_test() gains a scale
argument, to explicitely modulate the scale of the contrasts or
comparisons (e.g. "response" or "link", or
"exp" to return transformed
contrasts/comparisons).
hyothesis_test() now includes the response level for
models with ordinal outcomes (and alike).
When ggpredict() is used inside functions and a name
for a vector variable (passed as argument to that function) in
terms is used, the variable is now correctly
recognized.
Partial residuals (when plot(..., residuals = TRUE))
now supports more linear (mixed) models, including models from package
lme (such as gls() or
lme()).
For mixed models, type = "random" used to calculate
prediction intervals that always accounted for random effects
variances, leading to larger intervals. Using
interval = "confidence" together with
type = "random" now allows to calculate “usual” confidence
intervals for random effects. This is usefule for predictions at
specific group levels of random effects (when focal terms are only fixed
effects, use type = "fixed" for regular confidence
intervals).
The vcov.fun argument can now also be a function
that returns a variance-covariance matrix.
The verbose argument in ggpredict() and
hypothesis_test() now also toggle messages for the
respective print() methods.
The print() method for
hypothesis_test() has been revised and now provides more
details for possible transformation of the scale of comparisons and
contrasts.
The print() method now shows all rows by default
when the focal term is a factor. If rows are not shown in the output, a
message is printed to inform the user about truncated output.
A new vignette about using ggeffects in the context of an intersectional multilevel analysis of individual heterogeneity, using the MAIHDA framework.
Fixed issue with wrong order of x-axis-labels for plots when the focal term on the x-axis was a character vector, where alphabetical order of values did not match order of predictions.
Fixed issues in hyothesis_test() for models with
ordinal outcomes (and alike).
Added a new [.ggeffects function, which allows to
subset ggeffects objects in the same way as regular data
frames, i.e. it is now possible to do:
gge <- ggpredict(model, "x1")
gge[c(1:2)]Using a name for a vector variable in terms now
works from inside functions. E.g., you can now do:
foo <- function(data) {
fit <- lm(barthtot ~ c12hour + c172code, data = data)
v <- c(20, 50, 70)
ggpredict(fit, terms = "c12hour [v]")
}
foo(efc)The colors argument in plot() can now
also be applied to single-colored plots.
hyothesis_test() gains a collapse_level
argument, to collapse term labels that refer to the same levels into a
singel unique level string.
Fixed issue with misplaced residuals when x-axis was categorical and the factor levels were not in alphabetical order.
pool_predictions() now correctly handles models with
transformed response variables (like log(y)) and returns
the correct back-transformed pooled predictions (and their confidence
intervals).
Fixed issue with wrong computation of confidence intervals for
models of class clm from package ordinal.
Fixed failing tests due to changes in the logistf
package, which now also supports emmeans. That means,
ggemmeans() now also works for models from package
logistf.
Fixed bug in plot() when partial residuals were
added (i.e. residuals = TRUE) and
collapse.group was provided (in case of mixed
models).
Fixed issue with on-the-fly created factors inside formulas,
which were not correctly treated as factors in the plot()
method. This bug was related to recent changes in
insight::get_data().
Fixed issue with wrong labels in hyothesis_test()
for comparisons with many rows, when betas starting with same digit were
specified (e.g. test = "(b1-b13)=(b3-b15)").
Fixed issue in hyothesis_test() for mixed models
when focal terms included factors with factor levels that contained a
comma.
Fixed issue with missing confidence intervals for mixed models
when one of the variable names contains white space characters
(e.g. y ~ 'x a' + xb).
mblogit (mclogit),
phylolm and phyloglm (phylolm)
models.hypothesis_test() gains an equivalence
argument, to compute tests of practical equivalence for contrasts and
comparisons.
The message whether contrasts or comparisons from
hypothesis_test() are on the link-scale is now printed
below the table.
Dot arguments (...) in
hypothesis_test() are now passed to the functions in
marginaleffects, thereby allowing to use further options in
functions marginaleffects::predictions(), like
transform etc.
hypothesis_test() for mixed models with
one focal term only, and when this term was categorical.hypothesis_test(), to compute contrasts and
comparisons of predictions and test differences for statistical
significance. Additionally, an accompanying vignette that explains the
new function in detail is added.
install_latest(), to install the latest official
package version from CRAN, or the latest development version from
r-universe.
An as.data.frame() method was added, which converts
ggeffects objects returned by ggpredict() into
data frame, where standard column names are replaced by their related
variable names.
Response values are now also back-transformed when these were
transformed using log2(), log10() or
log1p().
The terms argument can now also be a named list.
Thus, instead of
terms = c("score [30,50,70]", "status [low, middle]") one
could also write
terms = list(score = c(30,50,70), status = c("low", "middle")).
Minor changes to meet forthcoming update of insight.
ggpredict() or ggemmeans() get a
verbose argument to suppress some messages and warnings
when calling
logitr (package logitr)Fixed issue with wrong standard errors for predicting random effect groups for more multiple levels.
Fixed issue in ggemmeans(), which did not correctly
averaged over character vectors when these were hold constant.
Fixed bug for models of class lme when
type = "re" was requested.
Fix wrong computations of predictions for
arm::bayesglm() models.
Fix CRAN check issues.
Speed improvement for some models when calculating uncertainty intervals of predictions.
Minor fixes.
mo() with numeric predictors, which only allow to predict
for values that are actually present in the data.Fixed issue with adding raw data points for plots from logistic regression models, when the response variable was no factor with numeric levels.
Fixed issues with CRAN checks.
orm (package rms)Prediction intervals (where possible, or when
type = "random"), are now always based on sigma^2
(i.e. insight::get_sigma(model)^2). This is in line with
interval = "prediction" for lm, or for predictions
based on simulations (when type = "simulate").
print() now uses the name of the focal variable as
column name (instead) of "x").
collapse_by_group(), to generate a data frame where the
response value of the raw data is averaged over the levels of a (random
effect) grouping factor.A new vignette was added related to the definition and meaning of “marginal effects” and “adjusted predictions”. To be more strict and to avoid confusion with the term “marginal effect”, which meaning may vary across fields, either “marginal effects” was replaced by “adjusted predictions”, or “adjusted predictions” was added as term throughout the package’s documentation and vignettes.
Allow confidence intervals when predictions are conditioned on
random effect groups (i.e. when type = "random" and
terms includes a random effect group factor).
Predicted response values based on simulate()
(i.e. when type = "simulate") is now possible for more
model classes (see ?ggpredict).
ggpredict() now computes confidence intervals for
some edge cases where it previously failed (e.g. some models that do not
compute standard errors for predictions, and where a factor was included
in the model and not the focal term).
plot() gains a collapse.group argument,
which - in conjunction with add.data - averages
(“collapses”) the raw data by the levels of the group factors (random
effects).
data_grid() was added as more common alias for
new_data().
ggpredict() and plot() for
survival-models now always start with time = 1.
Fixed issue in print() for survival-models.
Fixed issue with type = "simulate" for
glmmTMB models.
Fixed issue with gamlss models that had
random() function in the model formula.
Fixed issue with incorrect back-transformation of predictions for
geeglm models.
residuals.type argument in plot() is
deprecated. Always using "working" residuals.pretty_range() and values_at() can now
also be used as function factories.
plot() gains a limit.range argument, to
limit the range of the prediction bands to the range of the
data.
Fixed issue with unnecessary back-transformation of log-transformed offset-terms from glmmTMB models.
Fixed issues with plotting raw data when predictor on x-axis was a character vector.
Fixed issues from CRAN checks.
interval to ggemmeans(), to
either compute confidence or prediction intervals.averaging (package MuMIn)pool_predictions(), to pool multiple
ggeffects objects. This can be used when predicted values
or estimated marginal means are calculated for models fit to multiple
imputed datasets.residualize_over_grid() is now
exported.log1p() and
log(mu + x).type = "random" or
"zi_random"), but random effects variances could not be
calculated or were almost zero.multinom
models in ggemmeans().ggemmeans() for models from
nlme.plot() for some models in
ggeffect().terms = "predictor [exp]" is no longer necessary.mlogit (package mlogit)plot() now can also create partial residuals plots.
There, arguments residuals, residuals.type and
residuals.line were added to add partial residuals, the
type of residuals and a possible loess-fit regression line for the
residual data.glm since
some time. Should be fixed now.ggpredict() and rlmerMods
models when using factors as adjusted terms.mclogit (package mclogit)ggeffect().ggpredict() gets a new type-option,
"zi.prob", to predict the zero-inflation probability (for
models from pscl, glmmTMB and
GLMMadaptive).add.data = TRUE in plot(), the raw data points
are also transformed accordingly.plot() with add.data = TRUE first adds the
layer with raw data, then the points / lines for the marginal effects,
so raw data points to not overlay the predicted values.terms-argument now also accepts the name of a
variable to define specific values. See vignette Marginal Effects at
Specific Values.vcov.type was not specified.type-argument.1.offset() terms.mixor (package mixor),
cgam, cgamm (package
cgam)x.as.factor is considered as less useful
and was removed.fixest (package fixest),
glmx (package glmx).plot(rawdata = TRUE) now also works for objects from
ggemmeans().ggpredict() now computes confidence intervals for
predictions from geeglm models.trials() as response
variable, ggpredict() used to choose the median value of
trials were the response was hold constant. Now, you can use the
condition-argument to hold the number of trials constant at
different values.print().clmm-models, when group factor in
random effects was numeric.emm() is discouraged, and so it was
removed.bracl, brmultinom (package
brglm2) and models from packages
bamlss and R2BayesX.plot() now uses dodge-position for raw data for
categorical x-axis, to align raw data points with points and error bars
geoms from predictions.show_pals()).vcov() function to calculate
variance-covariance matrix for marginal effects.ggemmeans() now also accepts type = "re"
and type = "re.zi", to add random effects variances to
prediction intervals for mixed models.... is now passed down to the
predict()-method for gamlss-objects, so
predictions can be computed for sigma, nu and tau as well.ggeffect(), when one term was a character vector.ggaverage() is discouraged, and so it was
removed.rprs_values() is now deprecated, the function
is named values_at(), and its alias is
representative_values().x.as.factor-argument defaults to
TRUE.ggpredict() now supports cumulative link and ordinal
vglm models from package VGAM.terms included random effects.add.data is an alias for the
rawdata-argument in plot().ggpredict() and ggemmeans() now also
support predictions for gam models from ziplss
family.print()-method for ordinal or cumulative link
models.plot()-method no longer changes the order of factor
levels for groups and facets.pretty_data() gets a length() argument to
define the length of intervals to be returned.values_at() is an alias for
rprs_values().betabin, negbin (package
aod), wbm (package panelr)ggpredict() now supports prediction intervals for
models from MCMCglmm.ggpredict() gets a
back.transform-argument, to tranform predicted values from
log-transformed responses back to their original scale (the default
behaviour), or to allow predictions to remain on log-scale (new).ggpredict() and ggemmeans() now can
calculate marginal effects for specific values from up to three terms
(i.e. terms can be of lenght four now).ci.style-argument from plot() now also
applies to error bars for categorical variables on the x-axis.gamlss, geeglm (package
geepack), lmrob and glmrob
(package robustbase), ols (package
rms), rlmer (package
robustlmm), rq and rqss
(package quantreg), tobit (package
AER), survreg (package
survival)terms = "predictor [1:10]") can now be changed with
by, e.g. terms = "predictor [1:10 by=.5]" (see
also vignette Marginal Effects at Specific Values).vcov.fun in ggpredict()) now also works for
following model-objects: coxph, plm,
polr (and probably also lme and
gls, not tested yet).ggpredict() gets an interval-argument, to
compute prediction intervals instead of confidence intervals.plot.ggeffects() now allows different horizontal and
vertical jittering for rawdata when jitter is
a numeric vector of length two.AsIs-conversion from division of two
variables as dependent variable, e.g. I(amount/frequency),
now should work.ggpredict() failed for MixMod-objects when
ci.lvl=NA.