NEWS.md
tab_model()
now works properly with forthcoming parameters update.plot_models()
did not work properly for Bayesian models.tab_model()
also gains an encoding
argument.tab_df()
and tab_dfs()
no longer set the argument show.rownames
to TRUE
. Therefore, both functions now use row numbers as row names, if no other rownames are present.tab_dfs()
also gains a digits
argument.df.method
in tab_model()
did not accept all available options that were documented.dv.labels = ""
in tab_model()
, the row with names of dependent variables is omitted.minus.sign
argument in tab_model()
now works.show.std = TRUE
in tab_model()
did not exponentiate standardized coefficients for non-Gaussian models.tab_model()
gains an argument df.method
, which will replace the less generic p.val
argument in the future. Currently, df.method
is an alias of p.val
.plot_stackfrq()
when weights were applied.plot_stackfrq()
when weights were applied and items should be sorted.plot_models()
for models without intercept.plot_models()
when showing p-stars.plot_model()
with type = "int"
in detecting interaction terms when these were partly in parenthesis (like a * (b + c)
).tab_model()
with arguments show.stat = TRUE
and show.std = TRUE
, where the related statistic and CI columns for standardized coefficients were not shown.tab_model()
for brmsfit models that did no longer show random effects information after the last update from the performance package.show.rownames
in tab_df()
.vcov.fun
in tab_model()
or plot_model()
) now also uses and thus accepts estimation-types from package clubSandwich.tab_model()
now accepts all options for p.val
that are supported by parameters::model_parameters()
.p.style
argument in tab_model()
was slightly revised, and now also accepts "scientific"
as option for scientific notation of p-values.tab_model()
gets a digits.re
argument to define decimal part of the random effects summary.plot_models()
gains value.size
and line.size
arguments, similar to plot_model()
.plot_models()
should sort coefficients in their natural order now.plot_xtab()
with wrong order of legend labels.plot_models()
with wrong axis title for exponentiated coefficients.tab_model()
that did not show standard error of standardized coefficients when show.se = TRUE
.tab_model()
and plot_model()
now support clogit models (requires latest update of package insight).tab_model()
gets a p.adjust
argument to adjust p-values for multiple comparisons.tab_model()
, plot_model()
and plot_models()
get a robust
-argument to easily compute standard errors, confidence intervals and p-values based on robust estimation of the variance-covariance matrix. robust
is just a convenient shortcut for vcov.fun
and vcov.type
.tab_model()
and plot_model()
for certain cases when coefficients could not be estimated and were NA
.tab_model()
with collapse.ci
for Bayesian models.tab_model()
when p.val="kr"
and show.df=TRUE
.tab_model()
with formatting issues of p-values when standardized coefficients where requested.tab_model()
due to changes in other packages sjPlot depends on.sjt.itemanalysis()
is now named tab_itemscale()
.sjt.xtab()
is now named tab_xtab()
.tab_model()
of robust estimation in general and Kenward-Roger or Satterthwaite approximations in particular for linear mixed models.tab_df()
now uses value labels for factors instead of numeric values.tab_model()
gets arguments bootstrap
, iterations
and seed
to return bootstrapped estimates.tab_model()
with detecting labels when auto.label = TRUE
.tab_model()
for negative binomial hurdle mixed models (i.e. glmmTMB models with truncated negative-binomial family).tab_model()
with show.reflvl = TRUE
.tab_model()
where labels for coefficients where not matching the correct coefficients.plot_model()
or tab_model()
) now uses standardization based on refitting the model (see vignette for details).plot_model()
gets type = "emm"
as marginal effects plot type, which is similar to type = "eff"
. See Plotting Marginal Effects of Regression Models for details.verbose
-argument in view_df()
now defaults to FALSE
.show_pals()
).sort.est = NULL
in plot_model()
now preserves original order of coefficients.plot_frq()
for non-labelled, numeric values.plot_frq()
when plotting factors.string.std_ci
and string.std_se
are no longer ignored in tab_model()
.performance::principal_component()
by parameters::principal_component()
.sjp.grpfrq()
is now names plot_grpfrq()
.sjp.xtab()
is now names plot_xtab()
.plot_grid()
gets a tags
-argument to add tags to plot-panels.plot_stackfrq()
for data frames with many missing values.plot_frq()
when vector had more labels than values.tab_model()
where show.reflvl = TRUE
did not insert the reference category in first place, but in alphabetical order.show_sjplot_pals()
).tab_model()
now supports gamlss models.tab_df()
gets a digits
argument, to round numeric values in output.tab_model()
with show.df = TRUE
for lmerModLmerTest.tab_stackfrq()
when items had different amount of valid values.sjp.stackfrq()
was renamed to plot_stackfrq()
.sjt.stackfrq()
was renamed to tab_stackfrq()
.plot_likert()
group.legend.options
. The ordering now defaults to row wise and the user can force all categories onto a single row.tab_model()
wbm()
-models from the panelr-package.show.aicc
-argument to show the second order AIC.show.reflvl
-argument to show the reference level of factors.string.std_se
and string.std_ci
-argument to change the column header for standard errors and confidence intervals of standardized coefficients.show.ci50
defaults to FALSE
now.sjt.itemanalysis()
sjt.itemanalysis()
now works on ordered factors. A clearer error message was added when unordered factors are used. The old error message was not helpful.factor.groups
argument can now be "auto"
to detect factor groups based on a pca with Varimax rotation.sjp.stackrq()
sjp.stackfrq()
was renamed to plot_stackfrq()
.sjp.stackfrq()
(now named: plot_stackfrq()
) gets a show.n
-argument to also show count values. This option can be combined with show.prc
.sjp.stackfrq()
(now named: plot_stackfrq()
) now also works on grouped data frames.plot_model()
now supports wbm()
-models from the panelr-package.plot_model(type = "int")
now also recognized interaction terms with :
in formula.string.est
in tab_model()
did not overwrite the default label for the estimate-column-header.tab_model()
for mixed models that can’t compute R2.tab_model()
when printing robust standard errors and CI (i.e. when using arguments vcov*
).plot_likert()
option reverse.scale = TRUE
resulted in values = "sum.inside"
being outside and the other way around. This is fixed now.view_df()
mixed up labels and frequency values when value labels were present, but no such values were in the data.wrap.labels
in plot_frq()
did not properly work for factor levels.plot_models()
that stopped for some models.sjt.stackfrq()
, when show.na = TRUE
and some items had zero-values.dplyr::n()
, to meet changes in dplyr 0.8.0.plot_model()
and tab_model()
now support MixMod
-objects from package GLMMadpative, mlogit
- and gmnl
-models.sjp.kfold_cv()
was renamed to plot_kfold_cv()
.sjp.frq()
was renamed to plot_frq()
.tab_model()
gets a show.ngrps
-argument, which adds back the functionality to print the number of random effects groups for mixed models.tab_model()
gets a show.loglik
-argument, which adds back the functionality to print the model’s log-Likelihood.tab_model()
gets a strings
-argument, as convenient shortcut for setting column-header strings.tab_model()
gets additional arguments vcov.fun
, vcov.type
and vcov.args
that are passed down to sjstats::robust()
, to calculate different types of (clustered) robust standard errors.p.style
-argument now also allows printing both numeric p-values and asterisks, by using p.style = "both"
.plot_likert()
gets a reverse.scale
argument to reverse the order of categories, so positive and negative values switch position.plot_likert()
gets a groups
argument, to group items in the plot (thanks to @ndevln).grid.range
in plot_likert()
now may also be a vector of length 2, to define diffent length for the left and right x-axis scales.plot_frq()
(former sjp.frq()
) now has pipe-consistent syntax, enables plotting multiple variables in one function call and supports grouped data frames.plot_model()
gets additional arguments vcov.fun
, vcov.type
and vcov.args
that are passed down to sjstats::robust()
, to calculate different types of (clustered) robust standard errors.sjt.xtab()
, sjp.xtab()
, plot_frq()
and sjp.grpfrq()
get a drop.empty()
-argument, to drop values / factor levels with no observations from output.plot_model(..., type = "diag")
.color ="bw"
and legend.title
was specified.view_df()
did not truncate frequency- and percentage-values for variables where value labels were truncated to a certain maximum number.tab_model()
did not print number of observations for coxph
-models.Following functions are now defunct:
sjt.lm()
, sjt.glm()
, sjt.lmer()
and sjt.glmer()
. Please use tab_model()
instead.tab_model()
supports printing simplex parameters of monotonic effects of brms models.tab_model()
gets a prefix.labels
-argument to add a prefix to the labels of categorical terms.rotation
-argument in sjt.pca()
and sjp.pca()
now supports all rotations from psych::principal()
.plot_model()
no longer automatically changes the plot-type to "slope"
for models with only one predictor that is categorical and has more than two levels.type = "eff"
and type = "pred"
in plot_model()
did not work when terms
was not specified.tab_model()
, the confidence intervals and p-values are now re-calculated and adjusted based on the robust standard errors.colors = "bw"
was not recognized correctly for plot_model(..., type = "int")
.sjp.frq()
with correct axis labels for non-labelled character vectors.sjt.lm()
, sjt.glm()
, sjt.lmer()
and sjt.glmer()
are now deprecated. Please use tab_model()
instead.dot.size
and line.size
in plot_model()
now also apply to marginal effects and diagnostic plots.plot_model()
now uses a free x-axis scale in facets for models with zero-inflated part.plot_model()
now shows multiple plots for models with zero-inflated parts when grids = FALSE
.tab_model()
gets a p.style
and p.threshold
argument to indicate significance levels as asteriks, and to determine the threshold for which an estimate is considered as significant.plot_model()
and plot_models()
get a p.threshold
argument to determine the threshold for which an estimate is considered as significant.plot_likert()
.tab_model()
now also accepts multiple model-objects stored in a list
as argument, as stated in the help-file.file
-argument now works again in sjt.itemanalysis()
.show.ci
in tab_model()
did not compute confidence intervals for different levels.sjp.scatter()
was revised and renamed to plot_scatter()
. plot_scatter()
is pipe-friendly, and also works on grouped data frames.sjp.gpt()
was revised and renamed to plot_gpt()
. plot_gpt()
is pipe-friendly, and also works on grouped data frames.sjp.scatter()
was renamed to plot_scatter()
.sjp.likert()
was renamed to plot_likert()
.sjp.gpt()
was renamed to plot_gpt()
.sjp.resid()
was renamed to plot_residuals()
.brmsfit
-objects with categorical-family for plot_model()
and tab_model()
.tab_model()
gets a show.adj.icc
-argument, to also show the adjusted ICC for mixed models.tab_model()
gets a col.order
-argument, reorder the table columns.hide.progress
in view_df()
is deprecated. Please use verbose
now.statistics
-argument in sjt.xtab()
gets a "fisher"
-option, to force Fisher’s Exact Test to be used.Following functions are now defunct:
sjp.lm()
, sjp.glm()
, sjp.lmer()
, sjp.glmer()
and sjp.int()
. Please use plot_model()
instead.sjt.frq()
. Please use sjmisc::frq(out = "v")
instead.lmerModLmerTest
objects.show.std
) in tab_model()
.tab_model()
as replacement for sjt.lm()
, sjt.glm()
, sjt.lmer()
and sjt.glmer()
. Furthermore, tab_model()
is designed to work with the same model-objects as plot_model()
.scale_fill_sjplot()
and scale_color_sjplot()
. These provide predifined colour palettes from this package.show_sjplot_pals()
to show all predefined colour palettes provided by this package.sjplot_pal()
to return colour values of a specific palette.Following functions are now deprecated:
sjp.lm()
, sjp.glm()
, sjp.lmer()
, sjp.glmer()
and sjp.int()
. Please use plot_model()
instead.sjt.frq()
. Please use sjmisc::frq(out = "v")
instead.Following functions are now defunct:
sjt.grpmean()
, sjt.mwu()
and sjt.df()
. The replacements are sjstats::grpmean()
, sjstats::mwu()
and tab_df()
resp. tab_dfs()
.plot_model()
and plot_models()
get a prefix.labels
-argument, to prefix automatically retrieved term labels with either the related variable name or label.plot_model()
gets a show.zeroinf
-argument to show or hide the zero-inflation-part of models in the plot.plot_model()
gets a jitter
-argument to add some random variation to data points for those plot types that accept show.data = TRUE
.plot_model()
gets a legend.title
-argument to define the legend title for plots that display a legend.plot_model()
now passes more arguments in ...
down to ggeffects::plot()
for marginal effects plots.plot_model()
now plots the zero-inflated part of the model for brmsfit
-objects.plot_model()
now plots multivariate response models, i.e. models with multiple outcomes.plot_model()
(type = "diag"
) can now also be used with brmsfit
-objects.plot_model()
(type = "diag"
) for Stan-models (brmsfit
or stanreg
resp. stanfit
) can now be set with the axis.lim
-argument.grid.breaks
-argument for plot_model()
and plot_models()
now also takes a vector of values to directly define the grid breaks for the plot.plot_model()
and plot_models()
when the grid.breaks
-argument is of length one.terms
-argument for plot_model()
now also allows the specification of a range of numeric values in square brackets for marginal effects plots, e.g. terms = "age [30:50]"
or terms = "age [pretty]"
.terms
- and rm.terms
-arguments for plot_model()
now also allows specification of factor levels for categorical terms. Coefficients for the indicted factor levels are kept resp. removed (see ?plot_model
for details).plot_model()
now supports clmm
-objects (package ordinal).plot_model(type = "diag")
now also shows random-effects QQ-plots for glmmTMB
-models, and also plots random-effects QQ-plots for all random effects (if model has more than one random effect term).plot_model(type = "re")
now supports standard errors and confidence intervals for glmmTMB
-objects.glmmTMB
-tidier, which may have returned wrong data for zero-inflation part of model.brms
area now shown in each own facet per intercept.sjp.likert()
for uneven category count when neutral category is specified.plot_model(type = "int")
could not automatically select mdrt.values
properly for non-integer variables.sjp.grpfrq()
now correctly uses the complete space in facets when facet.grid = TRUE
.sjp.grpfrq(type = "boxplot")
did not correctly label the x-axis when one category had no elements in a vector.