insight 0.8.4 Unreleased

General

  • HRQoL was removed from suggested packages, as it was removed from CRAN.
  • Better support for dispersion models in glmmTMB.

Changes to functions

  • null_model() now also works for non-mixed models.
  • get_variance() now also computes variance components for models (from mixed models packages) without random effects.
  • Improved support for afex_aov and aovlist (i.e. Anova with error term).

Bug fixes

  • Fixed some issues with deparsings NULL strings under R 4.0.0.
  • Fixed accuracy in get_variance() for models from Gamma family.
  • Fixed edge case in clean_names().
  • Fixed issues with find_formula.lme() under R 4.0.0.
  • Fixed issues with examples from clean_names() under R-devel.

insight 0.8.3 2020-04-20

General

  • The function to calculate null-models for mixed effects models is now exported (null_model().)

New supported model classes

  • arima (stats), averaging (MuMIn)

Changes to functions

Bug fixes

  • Fixed issue in find_parameters() and clean_parameters() for brmsfit models with specific variable name patterns.
  • Fixed issue in format_ci() when confidence interval only contained NAs and width was set to "auto".
  • Fixed issue in find_formula() for mixed models when formula contained parentheses in the non-random parts, around a certain set of predictors.
  • Fixed issue in get_priors.BFBayesFactor() for BFMetat class.
  • Fixed issue in clean_parameters.BFBayesFactor() when model contained interaction terms and these were assigned to the “extra” component.

insight 0.8.2 2020-03-06

Breaking changes

  • model_info() now only returns TRUE for $is_ordinal, when model is an ordinal or cumulative link model. In past versions, $is_ordinal was also TRUE for multinomial models.

New supported model classes

  • bife (bife), bcplm and zcpglm (cplm)

General

  • Improved support for clogit-models.

Bug fixes

  • Fixed issue in find_weights() for merMod models.
  • Fixed issue in get_data() for models with weights, when weights also contained missing data.
  • Fixed issue in get_data() for mixed models with complex offset-terms.
  • Fixed issue in get_statistic() for zeroinfl models with theta-coefficients.
  • Fixed issue in get_statistic() for lmerModLmerTest models with.
  • Fixed issue in find_parameters() for brmsfit models for rare situations where a specific pattern of variables names, when used as random effects, did not properly separate fixed from random effects in the return value.
  • Fixed issue related to CRAN checks.

insight 0.8.1 2020-02-02

New supported model classes

  • cglm (cglm), DirichletRegModel (DirichletReg).

General

Changes to functions

  • model_info() now returns $is_multinomial for multinomial (but not ordinal or cumulative) link models.
  • format_value() gets an as_percent argument to format values as percentages.

Bug fixes

  • Fixed issue in get_data() for clmm2-models.
  • Fixed issue in get_data() for models that used the lspline()-function.
  • Fixed issue in get_statistic() for multinom models.
  • Fixed issue in get_priors() for stanreg models with flat intercept priors.
  • Fixed tests that failed due to latest fixest update.

insight 0.8.0 2020-01-14

New supported model classes

  • brglm (brglm), cgam, cgamm (cgam), cpglm, cpglmm (cplm), feglm (apaca), glmmadmb (glmmADMB), glmx (glmx), partial support for mcmc (coda), mixor (mixor), MANOVA, RM (MANOVA.RM).

General

  • Better handling of clm2, clmm2 and rqss models.

New functions

  • format_ci() (re-implemented and slightly enhanced from parameters), to format confidence/credible intervals.

Changes to functions

  • find_parameters() now also works for BFBayesFactor objects.
  • Suppress non-informative warning in get_data() for model data with weights.
  • format_value() automatically uses scientific notation for very large numbers (> 1e+5). Furthermore, the check for integer values was made more robust, to avoid warnings when checking very large numbers for integer type.
  • Improved find_parameters(), get_parameters() and clean_parameters() for BFBayesFactor-objects.
  • get_priors() now works for stanmvreg objects.
  • Other minor improvements.

Bug fixes

  • Better detect Tweedie-models in model_info().
  • Fixed issue in find_random_slopes() for panelr-models with multiple random-effect parts.
  • Fixed issues with zerotrunc models.
  • Fixed issues with brmsfit models with correlated random effects.
  • Fixed issue with edge-cases in clean_names().
  • Fixed issue with breaking changes with latest brms-update.
  • Further minor bug fixes.

insight 0.7.1 2019-11-28

New supported model classes

  • complmrob (complmrob), fixest (fixest), mclogit and mmclogit (mclogit).

Bug fixes

  • Fixed bug in find_formula() for mixed models, when random effects are written before any fixed effects terms (like social ~ (1|school) + open + extro).
  • Fixed bug in model_info() for VGAM models, where logit-link was not always correctly identified.
  • Fixed issue in get_priors() for brmsfit models, where parameters of conditional and zero-inflated model components had identical names. This caused errors in bayestestR::simulate_prior().
  • Fixed CRAN check issue.

insight 0.7.0 2019-11-14

Breaking changes

  • In order to unify column names across easystats-packages, get_parameters() and get_priors() now return column names according to our naming conventions (i.e. capitalized column names).
  • model_info() returned both $is_zeroinf and $is_zero_inflated for zero-inflated models. Now $is_zeroinf is softly deprecated, so model_info() will return $is_zero_inflated only in future updates.

New supported model classes

  • aareg (survival), brmultinom and bracl (brglm2), and wbgee (panelr). Furthermore, for different model-types from panelr models (within-between, between, etc.) are now also supported.
  • Preliminary support for rma models (metafor).

Changes to functions

  • get_statistic() supports multinom models (nnet).
  • link_inverse() gets a what-argument, to return the link-inverse function for specific distribution parameters from gamls models.

Bug fixes

insight 0.6.0 2019-10-17

New supported model classes

New functions

Changes to functions

  • model_info() now returns the element is_survival for survival models.
  • model_info() now returns the element is_truncated for truncated regression, or brmsfit models with trunc() as additional response part.
  • model_info() now recognizes beta and beta inflated families from package gamlss.
  • Better support for nonlinear quantile regression (quantreg::nlrq()).
  • Better support for nonlinear mixed models (lme4::nlmer()). Note that model-specification requires the random term to be written in parentheses, i.e. (slope | group).

Bug fixes

insight 0.5.0 2019-09-07

Breaking Changes

  • The previous is_model() function has been renamed to is_model_supported() since it was unclear if the function checked the entered object was a model or a supported model in insight. The new is_model() function checks if the entered object is a model object, while is_model_supported() checks if a supported model object.

New functions

  • find_statistic() to return the test statistic of a regression model.
  • format_value() and format_table() as utility-functions to format (model) output, especially for tabular output.
  • color_if() as utility-function to add color formatting to values, depending on certain conditions.

General

  • Make extraction of model family information more stable for gam-objects.

Changes to functions

Bug fixes

  • find_formula() now extracts group factors of random effects for gamlss models.
  • find_parameters() and get_parameters() no longer show NA coefficients from group factors of random effects for gamlss models.
  • find_parameters() and get_parameters() did not work for multivariate response models of class brmsfit when argument parameters was specified.
  • get_data() dropped value and variable label attributes, when model frame contained matrix variables (like splines).
  • get_priors() swapped column names location and scale for brmsfit -objects.
  • get_parameters() did not work for glmmTMB models without zero-inflation component.
  • find_predictors() did not remove parentheses from terms in multiple nested random effects.
  • Better support for gam models (package mgcv) with ziplss or mvn families.

insight 0.4.1 2019-07-24

Changes to functions

Bug fixes

  • Fix CRAN-check issues due to recent update from the panelr-package.