:bar_chart: Computation and processing of models' parameters
serp (serp).include_reference can now directly be set to TRUE in model_parameters()
and doesn't require a call to print() anymore.
compare_parameters() gains a include_reference argument, to add the
reference category of categorical predictors to the parameters table.
print_md() for compare_parameters() now by default uses the tinytable
package to create markdown tables. This allows better control for column
heading spanning over multiple columns.
Fixed issue with parameter names for model_parameters() and objects from
package epiR.
Fixed issue with exponentiate = TRUE for model_parameters() with models
of class clmm (package ordinal), when model had no component column
(e.g., no scale or location parameters were returned).
include_reference now also works when factor were created "on-the-fly" inside
the model formula (i.e. y ~ as.factor(x)).
exponentiate argument of model_parameters() for
marginaleffects::predictions() now defaults to FALSE, in line with all
the other model_parameters() methods.model_parameters() for models of package survey now gives informative
messages when bootstrap = TRUE (which is currently not supported).
n_factors() now also returns the explained variance for the number of
factors as attributes.
model_parameters() for objects of package metafor now warns when unsupported
arguments (like vcov) are used.
Improved documentation for pool_parameters().
print(include_reference = TRUE) for model_parameters() did not work when
run inside a pipe-chain.
Fixed issues with format() for objects returned by compare_parameters()
that included mixed models.
principal_components() and factor_analysis() now also work when argument
n = 1.
print_md() for compare_parameters() now gains more arguments, similar to
the print() method.
bootstrap_parameters() and model_parameters() now accept bootstrapped
samples returned by bootstrap_model().
The print() method for model_parameters() now also yields a warning for
models with logit-links when possible issues with (quasi) complete separation
occur.
Fixed issue in print_html() for objects from package ggeffects.
Fixed issues for nnet::multinom() with wide-format response variables (using
cbind()).
Minor fixes for print_html() method for model_parameters().
Robust standard errors (argument vcov) now works for plm models.
Minor improvements to factor analysis functions.
The ci_digits argument of the print() method for model_parameters() now
defaults to the same value of digits.
model_parameters() for objects from package marginaleffects now also
accepts the exponentiate argument.
The print(), print_html(), print_md() and format() methods for
model_parameters() get an include_reference argument, to add the reference
category of categorical predictors to the parameters table.
Fixed issue with wrong calculation of test-statistic and p-values in
model_parameters() for fixest models.
Fixed issue with wrong column header for glm models with
family = binomial("identiy").
Minor fixes for dominance_analysis().
nestedLogit (nestedLogit).model_parameters() now also prints correct "pretty names" when predictors
where converted to ordered factors inside formulas, e.g. y ~ as.ordered(x).
model_parameters() now prints a message when the vcov argument is provided
and ci_method is explicitly set to "profile". Else, when vcov is not
NULL and ci_method is NULL, it defaults to "wald", to return confidence
intervals based on robust standard errors.
lme.mipo for models with ordinal or
categorical outcome.Added support for models of class hglm (hglm), mblogit (mclogit),
fixest_multi (fixest), and phylolm / phyloglm (phylolm).
as.data.frame methods for extracting posterior draws via bootstrap_model()
have been retired. Instead, directly using bootstrap_model() is recommended.
equivalence_test() gets a method for ggeffects objects from package
ggeffects.
equivalence_test() now prints the SGPV column instead of % in ROPE.
This is because the former % in ROPE actually was equivalent to the second
generation p-value (SGPV) and refers to the proportion of the range of the
confidence interval that is covered by the ROPE. However, % in ROPE did
not refer to the probability mass of the underlying distribution of a confidence
interval that was covered by the ROPE, hence the old column name was a bit
misleading.
Fixed issue in model_parameters.ggeffects() to address forthcoming changes
in the ggeffects package.
When an invalid or not supported value for the p_adjust argument in
model_parameters() is provided, the valid options were not shown in correct
capital letters, where appropriate.
Fixed bug in cluster_analysis() for include_factors = TRUE.
Fixed warning in model_parameters() and ci() for models from package
glmmTMB when ci_method was either "profile" or "uniroot".
Reduce unnecessary warnings.
The deprecated argument df_method in model_parameters()was removed.
Output from model_parameters() for objects returned by manova() and
car::Manova() is now more consistent.
Fixed issues in tests for mmrm models.
Fixed issue in bootstrap_model() for models of class glmmTMB with
dispersion parameters.
Fixed failing examples.
flic and flac (logistf), mmrm (mmrm).model_parameters() now includes a Group column for stanreg or brmsfit
models with random effects.
The print() method for model_parameters() now uses the same pattern to
print random effect variances for Bayesian models as for frequentist models.
Fixed issue with the print() method for compare_parameters(), which
duplicated random effects parameters rows in some edge cases.
Fixed issue with the print() method for compare_parameters(), which
didn't work properly when ci=NULL.