version 0.4.0

Breaking changes

• requires R >= 4.3.0 and the reformulas package; BayesTools now compiles a JAGS module (inverse-gamma, moment, inverse-moment, and LKJ distributions) and requires JAGS 4.x (>= 4.3.0), including its headers and libraries for installation from source
• fits, mixed and marginal posteriors, marginal likelihoods, and weight-function priors created with BayesTools 0.3.0 or earlier (including RoBMA 4.0.0 models) must be refitted or recreated; post-fit functions stop on them with a BayesTools_refit_required error
• marginal likelihoods are BayesTools_marglik objects (still inheriting from "bridge"), and the model-averaging functions accept only these; use JAGS_bridgesampling() or bridgesampling_object(logml)
• inverse-gamma priors use the module distribution dbt_invgamma on the parameter itself instead of a gamma prior on inv_<parameter>
• removes the deprecated transform_orthonormal argument (use transform_factors) and the ineffective seed argument of JAGS_extend(); pass the arguments that followed them by name
• check_bool() rejects NA by default and check_int() rejects infinite values
• fixed weight-function priors require a reference weight of exactly 1, and cumulative weight-function and Dirichlet priors require concentrations of at least 0.01
• stricter inputs: finite autofit convergence thresholds, named add_bounds, factor priors outside formulas need prior_factor_levels(), formula random effects need prior_random(), and mean-difference or orthonormal priors on indicator-coded interaction terms (e.g., g:x in ~ g + g:x) and multiply_by on the intercept prior or on a component of a mixture or spike-and-slab term prior (where it was ignored) are rejected, and interaction priors that are not invariant to the mean-difference or orthonormal basis of a factor (anything but zero-mean normal and zero-spike priors, or symmetric priors for two-level factors) are rejected
• JAGS_evaluate_formula(), transform_scale_samples(), and original-scale transformations require the formula design stored by JAGS_fit(); attach it to draws without a fit with JAGS_formula_draws()
• chain seeds are drawn from seed with R’s random-number generator instead of seed + chain, and mix_posteriors() allocates draws with multinomial counts, so seeded results differ from 0.3.0; seeded functions restore the caller’s random-number state instead of resetting it
• contr.orthonormal() and contr.meandif() return a closed-form (normalised Helmert) basis identical on every platform, so raw orthonormal and mean-difference coefficients differ from 0.3.0
• conditioning on an unknown parameter name is an error instead of a warning
• Savage_Dickey_BF() requires an exact, finite, and nonzero prior ordinate at the null hypothesis (classed errors otherwise), returns a finite Bayes factor when the null lies outside the posterior draws, and returns NA with the reason for levels with a posterior point mass at the null
• JAGS_check_convergence() and autofit classify parameters by their declared roles: constant parameters are structural, sampled parameters with constant draws fail unless allow_not_assessable = TRUE, and indicators are identified from their priors rather than their names
• parameter labels are rendered from structured label parts, which changes some labels: contrast coefficients that are not levels are labelled term{j}, interaction cells and formula prefixes are named consistently across tables, Bayes factor MC-error warnings name parameters by their table labels, and individual weight-function omegas in plots follow the order of summary tables
• JAGS_fit() monitors point priors
• posterior draw metadata are stored in one validated attribute (posterior_metadata()); arithmetic drops the metadata, values modified in place stop with BayesTools_stale_metadata, and posterior plots take point masses only from declared posterior atoms

Features

• adds formula random effects via prior_random(), random_block(), and related helpers, with lme4-like syntax (independent, nested, and interacted grouping, factor random slopes, where a mean-difference or orthonormal factor term is one exchangeable unit with one SD) and diagonal, unstructured (LKJ prior), compound-symmetry, AR(1), heterogeneous, and continuous-time AR(1) covariance structures
• adds random-effect variance allocations with Dirichlet weights and optional inclusion gates (random_variance_allocation()), known group covariance kernels (random_group_covariance()), external SD sources (random_sd_source()), spike-and-slab and mixture priors on random-effect SDs and correlations, and noncentered, centered, or mean-centered (hierarchically centered) parameterizations
• adds predictions with random effects (JAGS_predict_formula(); formula_target, blocks, and new_levels in JAGS_evaluate_formula()), marginal covariance helpers, and exact marginalization of Gaussian random-effect blocks in JAGS_bridgesampling()
• adds random-effect summaries to estimates tables (random_effects_summary and related filters) on the fitted or original scale, with level-cell correlations and multiple correlations (cor_total) of factor terms, and random_effects_summary_posterior()
• adds the RandomEffects vignette comparing BayesTools formula random effects with lme4 and brms
• adds prior_ordered() for ordered-factor priors that split a total effect across level increments with fixed or Dirichlet allocations
• adds Dirichlet simplex priors and moment and inverse-moment nonlocal priors
• adds prior_factor_levels() for attaching factor metadata to factor priors
• adds selection_model() for specifying how random effects and sampling errors enter selection models, and selection helpers for downstream packages
• adds a versioned parameter map stored with each fit, with parameter_catalog(), parameter_catalog_resolve(), parameter_draws(), parameter_labels(), and parameter_mixed_posterior() for selecting, extracting, and labelling model quantities
• adds JAGS_deterministic_nodes() and JAGS_evaluate_deterministic() for the deterministic quantities generated in JAGS models, and parameter_source() for referencing existing JAGS nodes
• adds posterior_metadata(), metadata constructors, and posterior_transform() for transforming posterior draws together with their supports, atoms, and densities
• adds parameter_prior_density(), prior_density_ordinate(), and prior_ordinate_status() for exact prior densities and ordinates of model quantities, including random-effect SDs, variance allocations, and LKJ correlations
• adds hypothesis_BF() for point and region hypotheses specified as expressions (including comparisons of factor levels), with the hypothesis parser hypothesis_parse() and hypothesis_linear_target()
• hypothesis_BF() also tests the quantities of a fitted model by the names of its parameter catalog
• adds hypothesis_BF_model(), hypothesis_provider(), and register_hypothesis_density_backend() for downstream model classes and density methods
• adds the density_method argument to Savage_Dickey_BF(), plot_posterior(), and plot_marginal() for posterior densities precomputed by downstream packages
• adds JAGS_formula_coefficient_transform() and JAGS_formula_prior_density() for original-scale prior densities of scaled formula coefficients
• adds JAGS_formula_draws() for evaluating formulas on draws without a fit
• adds runtime_setup, runtime_cache, and worker_output to JAGS_fit() and JAGS_extend(), jags_modules to JAGS_fit(), and the parallel cluster helpers JAGS_runtime_cluster() and JAGS_runtime_cluster_stop()
• adds monitor and allow_not_assessable to JAGS_check_convergence() and autofit_control, with per-parameter convergence diagnostics
• adds seed, repetitions, method, cores, and nonfinite to JAGS_bridgesampling(), and JAGS_marglik_priors_rows() for row-wise prior densities
• adds on_failure to the model-averaging functions for models whose marginal likelihood failed
• adds harrell_davis_quantile(); posterior PET-PEESE bands use Harrell-Davis quantiles
• adds as_marginal_inference(compute_BF = FALSE)
• exports helpers used by downstream packages (Bayes factor formatting, formula scaling, random-effect grouping, and posterior metadata helpers)

Changes

• JAGS_estimates_table() uses transform_scaled for both fixed and random effects; original-scale rows that cannot be computed are reported as NA with a note
• marginal likelihoods exceeding the iteration limit keep their estimate and warn
• JAGS_extend() recompiles the model from the stored chain states, so repeated extensions and extensions of reloaded fits are reproducible
• estimates tables identify inclusion rows from the priors instead of the row labels
• small Bayes factor bounds print as reciprocals (e.g., <1/10000)

Fixes

• prior densities of linear combinations (formula levels, marginal means, original-scale coefficients, and Savage-Dickey ordinates) use exact routes where available instead of a numerical grid, which was off by up to 20% for multiply_by products
• mean() and var() of truncated normal priors are computed analytically; numerical integration was wrong far in the tails and for narrow truncations
• far-tail truncated normal, t, and Cauchy priors sample and invert exactly
• the Savage-Dickey posterior ordinate is the exact kernel density at the null, reflected at exact support bounds; the previous grid estimate overestimated the density of long-tailed posteriors
• original-scale transformations of formulas such as ~ g + g:x and ~ f/x are derived from the fitted design; formulas without an original-scale representation stop
• prior densities of monitored formula coefficients no longer apply multiply_by scaling
• expression() formula terms are included in marginal-likelihood reconstruction
• log(intercept) formulas in marginal_posterior() and as_mixed_posteriors(transform_scaled = TRUE)
• marginal_posterior() for intercept-only formulas, predictors entering only through interactions, and model averages in which a model omits a factor term
• naming of interaction coefficients without main effects (e.g., g:x in ~ g + g:x)
• formulas generating duplicate JAGS node names stop with an informative error
• plot_posterior() bias prior overlays follow the full conditioning of the samples
• prior overlays weight spike-and-slab priors by the inclusion probability instead of 50/50
• individual omega and weight-function posterior plots with duplicate or zero-weight bias priors
• factor plot labels of interaction cells and secondary axes of mixed density plots
• marginal_estimates_table() reports Bayes factor warnings of scalar parameters
• stan_estimates_table() handles matrix-valued parameters
• p-hacking null masses no longer become negative for far-tail cut points
• JAGS_bridgesampling() stops on rank-deficient bridge coordinates instead of returning a meaningless estimate
• package contrasts resolve without attaching BayesTools

Performance

• JAGS_fit() resolves the JAGS executable once per session (about 0.8 s faster per fit)
• prior density curves and PET-PEESE prior bands are evaluated in batches (prior band plots take seconds instead of minutes)
• JAGS_bridgesampling() compiles formula and prior evaluators once per call
• two-bin cumulative weight-function priors are sampled through their Beta marginal
• saved fits no longer contain the compiled rjags model
• as_mixed_posteriors(transform_scaled = TRUE) caches the transformed prior densities

Development

• vignettes generate missing or outdated fitted-model caches when knitted; set BAYESTOOLS_VIGNETTE_REFIT=false to render without fitting
• declares waldo as a test dependency

version 0.3.0

Features

• major refactoring and speed-up of unit tests
• adds support for __default_factor and __default_continuous priors in JAGS_formula() - when specified in the prior_list, these are used as default priors for factor and continuous predictors that are not explicitly specified
• adds automatic standardization of continuous predictors via formula_scale parameter in JAGS_formula() and JAGS_fit() - improves MCMC sampling efficiency and numerical stability
• adds transform_scale_samples() function to transform posterior samples back to original scale after standardization
• adds transform_prior_samples() function to generate and transform prior samples using the same matrix transformation as posterior samples - enables correct visualization of priors on the original (unscaled) predictor scale, including proper handling of the intercept which depends on multiple coefficient priors
• adds transform_scaled argument to plot_posterior() for visualizing prior and posterior distributions on the original (unscaled) scale when using formula-based models with auto-scaling
• adds exp_lin transformation type for log-intercept unscaling in density/plotting functions: exp(a + b * log(x))
• adds log(intercept) formula attribute for specifying models of the form log(intercept) + sum(beta_i * x_i) - useful for parameters that must be positive (e.g., standard deviation) while keeping the intercept on the original scale. Set via attr(formula, "log(intercept)") <- TRUE. Supported in JAGS_formula(), JAGS_evaluate_formula(), and marginal likelihood computation
• adds advanced parameter filtering options to runjags_estimates_table():
  • remove_parameters = TRUE to remove all non-formula parameters
  • remove_formulas to remove all parameters from specific formulas
  • keep_parameters to keep only specified parameters
  • keep_formulas to keep only parameters from specified formulas
  • when bias is specified in remove_parameters or keep_parameters, the corresponding bias-related parameters (PET, PEESE, omega, alpha, pi_null, and phack_kind) are automatically included based on the bias prior type
• adds probs argument to runjags_estimates_table() and runjags_estimates_empty_table() for custom quantiles (default: c(0.025, 0.5, 0.975))
• adds effect_direction argument to plot_posterior(), plot_prior_list(), lines_prior_list(), and geom_prior_list() for PET-PEESE regression plots - use "positive" (default) for mu + PET*se + PEESE*se^2 or "negative" for mu - PET*se - PEESE*se^2
• redesigns prior_weightfunction() around a unified side, steps, and weights specification, with wf_cumulative(), wf_fixed(), and wf_independent() constructors for cumulative Dirichlet, fixed, independent, and log-independent weightfunction priors
• adds p-hacking and composed selection-bias priors via prior_phacking(), prior_bias(), calibration helpers, and selection_backend_spec() for compiling active step/p-hacking backend parameters
• adds error % for inclusion BF calculation

Changes

• changes quantile column names in runjags_estimates_table() and stan_estimates_table() from lCI/Median/uCI to numeric values (e.g., 0.025/0.5/0.975) for consistency with ensemble summary tables
• implied prior distributions for estimated marginal means, unstandardized coefficients, and PET-PEESE no longer require prior samples
• implied prior distributions for weightfunction weights now use analytical forms for cumulative Dirichlet, fixed, independent, and log-independent priors, including mixture and model-averaged weightfunctions where possible
• independent weightfunction priors now allow non-reference weights above one via non-negative omega-scale priors or unrestricted log-omega priors
• replaces the legacy dot-named weightfunction prior specifications with the unified weightfunction prior API and updates JAGS generation, marginal likelihood computation, posterior extraction, diagnostics, and summary tables to use the new component-local omega representation
• composed selection-bias priors and publication-bias mixtures now support prior sampling and explicit unsupported-operation errors for ambiguous scalar prior generics

Fixes

• reports inclusion Bayes factors as NA when the prior assigns probability 0 or 1 to inclusion, while keeping finite-sample bounds for posterior inclusion probabilities of 0 or 1
• fixes incorrect ordering the printed mixture priors
• fixes formula with no intercepts coded as 0 (instead of only -1)
• fixes bug in .is.wholenumber with NAs and na.rm = TRUE
• fixes ggplot prior spike layers for marginal factor plots with density and point components

version 0.2.23

Fixes

• JAGS_diagnostics functions now correctly handle factor parameters nested within mixture priors

version 0.2.22

Fixes

• plot_posterior() function with spike and slab priors

Changes

• unifies back-end of prior_mixture() and prior_spike_and_slab()

version 0.2.21

Fixes

• JAGS_formula() function now replaces removed missing intercept with 0 (so the model matrix remains unchanged)
• resetting silent = FALSE argument in the JAGS_fit() function now fits the model non-silently again

version 0.2.20

Features

• extending prior functions to accept expression() instead of a parameter, such objects can be use to create prior distributions that depend on other parameters in JAGS
• extending the formula interface of JAGS_fit() function to accept expressions that are appended as literal text to the generated JAGS formula
• extending the formula interface of JAGS_fit() function to handle uncorrelated random effects via (x||y) (lme4-like) notation

Fixes

• JAGS_estimates_table not printing formula prefix when only spike and slab priors are supplied

version 0.2.19

Features

• adds max_extend option to autofit_control argument in JAGS_fit() to limit the number of iterations for the model extension
• adds JASP progress bar integration

Fixes

• JAGS_diagnostics_density() plots for mixture distributions
• prior and posterior plot_posterior() for simple as_mixed_posteriors objects
• JAGS_evaluate_formula() for mixture and spike and slab priors
• set Bayes factors based on alternative only prior distributions to NA
• better handling of posterior samples in .fit_to_posterior()

version 0.2.18

Features

• adding prior_mixture() function for creating a mixture of prior distributions
• adding as_mixed_posteriors() and as_marginal_inference() functions for a single JAGS models (with spike and slab or mixture priors) to enabling tables and figures based on the corresponding output
• adding interpret2() function for another way of creating textual summaries without the need of inference and samples objects
• speedup and improvements to the runjags_estimates_table() function

Fixes

• small fixes for expansion of the RoBMA functionality

version 0.2.17

Features

• adding informed prior distributions for dichotomous and time to event outcomes based on Cochrane Database of Systematic Reviews to prior_informed() function
• adding bridge object convenience function bridge_object() (fixes: https://github.com/FBartos/BayesTools/issues/28)
• adding Na/NaN tests for check_ functions (fixes: https://github.com/FBartos/BayesTools/issues/26)

Fixes

• ability to run more than 4 chains (fixes: https://github.com/FBartos/BayesTools/issues/20)

version 0.2.16

Features

• update an existing JAGS fit with JAGS_extend() function
• new element of the autofit_control argument in JAGS_fit(): "restarts" allows to restart model initialization up to restarts times in case of failure

version 0.2.15

Fixes

• fixing repeated print of previous prior distribution in model_summary_table() in case of prior_none()

version 0.2.14

Features

• adding contrast = "meandif" to the prior_factor function which generates identical prior distributions for difference between the grand mean and each factor level
• adding contrast = "independent" to the prior_factor function which generates independent identical prior distributions for each factor level
• remove_column function for removing columns from BayesTools_table objects without breaking the attributes etc…
• adding empty table functions (https://github.com/FBartos/BayesTools/issues/10)
• adding remove_parameters argument to model_summary_table()
• adding multivariate point distribution functions
• adding point prior distribution as option to prior_factor with "meandif" and "orthonormal" contrasts
• adding marginal_posterior() function which creates marginal prior and posterior distributions (according to a model formula specification)
• adding Savage_Dickey_BF() function to compute density ratio Bayes factors based on marginal_posterior objects
• adding marginal_inference() function to combine information from marginal_posterior() and Savage_Dickey_BF()
• adding marginal_estimates_table() function to summarize marginal_inference() objects
• adding plot_marginal() function to visualize marginal_inference() objects

Changes

• contrast = "meandif" is now the default setting for prior_factor function
• depreciating transform_orthonormal argument in favor of more general transform_factors argument
• switching dummy contrast/factor attributes to treatment for consistency (https://github.com/FBartos/BayesTools/issues/23)

Fixes

• zero length inputs to check_bool(), check_char(), check_real(), check_int(), and check_list() do not throw error if allow_NULL = TRUE
• properly aggregating identical priors in the plotting function (previously overlying multiple spikes on top of each other when attributes did not match)
• student-t allowed as a prior distribution name
• fixing factor contrast settings in JAGS_evaluate_formula
• fixing spike prior transformations

version 0.2.13

Features

• runjags_estimates_table() function can now handle factor transformations
• plot_posterior function can now handle factor transformations
• ability to remove parameters from the runjags_estimates_table() function via the remove_parameters argument

Fixes

• inability to deal with constant intercept in marglik formula calculation
• runjags_estimates_table() function can now remove factor spike prior distributions
• marginal likelihood calculation for factor prior distributions with spike
• mixing samples from vector priors of length 1
• same prior distributions not always combined together properly when part of them was generated via the formula interface

version 0.2.12

Features

• stan_estimates_summary() function
• reducing dependency on runjags/rjags

Fixes

• dealing with posterior samples from rstan
• dealing with vector posterior samples
• fixing MCMC error of SD calculation for transformed samples (previously reported 100 times lower)

version 0.2.11

Features

• adding Bernoulli prior distribution
• adding spike and slab type of prior distributions (without marginal likelihood computations/model-averaging capabilities)
• new vignette comparing Bayes factor computation via marginal likelihood and spike and slab priors

Fixes

• when a transformation is applied, JAGS summary tables now produce the mean of the transformed variable (previous versions incorrectly returned transformation of the mean)

Changes

• runjags_XXX_table functions are now also exported as JAGS_XXX_functions for consistency with the rest of the code

version 0.2.10

Features

• trace, density, and autocorrelation diagnostic plots for JAGS models

version 0.2.9

Fixes

• dealing with NaNs in inclusion Bayes factors due to overflow with very large marginal likelihoods

version 0.2.8

Fixes

• dealing with point prior distributions in JAGS_marglik_parameters_formula function
• posterior samples dropping name in runjags_estimates_table function
• ensemble_summary_table and ensemble_diagnostics_table function can create table without model components

version 0.2.7

Features

• JAGS_evaluate_formula for evaluating formulas based on data and posterior samples (for creating predictions etc)
  
• JAGS_parameter_names for transforming formula names into the JAGS syntax

version 0.2.6

Features

• plot_models implementation for factor predictors
• format_parameter_names for cleaning parameter names from JAGS
• mean, sd, and var functions now return the corresponding values for differences from the mean for the orthonormal prior distributions

Fixes

• proper splitting of transformed posterior samples based on orthonormal contrasts in runjags_summary_table function (previous version crashed under other than default fit_JAGS settings)
• always showing name of the comparison group for treatment contrasts in runjags_summary_table function
• better handling of transformed parameter names in plot_models function

version 0.2.5

Features

• add_column function for extending BayesTools_table objects without breaking the attributes etc…
• ability to suppress the formula parameter prefix in BayesTools_table functions with with formula_prefix argument

Fixes

• allowing to pass point prior distributions for factor type predictors

version 0.2.4

Features

• adding possibility to multiply a (formula) prior parameter by another term (via multiply_by attribute passed with the prior)
• t-test example vignette

version 0.2.3

Fixes

• fixing error from trying to rename formula parameters in BayesTools tables when multiple parameters were nested within a component

version 0.2.2

Fixes

• fixing layering of prior and posterior plots in plot_posterior (posterior is now plotted over the prior)

version 0.2.1

Fixes

• fixing JAGS code for multivariate-t prior distribution

version 0.2.0

Changes

• ensemble inference, summary, and plot functions now extract the prior list from attribute of the fit objects (previously, the prior_list needed to be passed for each model within the model_list as the priors argument

Features

• adding formula interface for fitting and computing marginal likelihood of JAGS models
• adding factor prior distributions (with treatment and orthonormal contrasts)

version 0.1.4

Fixes

• fixing DOIs in the references file
• adds marglik argument inclusion_BF to deal with over/underflow (Issue #9)
• better passing of BF names through the ensemble_inference_table() (Issue #11)

Features

• adding logBF and BF01 options to ensemble_summary_table (Issue #7)

version 0.1.3

Features

• prior_informed function for creating informed prior distributions based on the past psychological and medical research

version 0.1.2

Fixes

• prior.plot can’t plot “spike” with plot_type == "ggplot" (Issue #6)
• MCMC error/SD print names in BayesTools tables (Issue #8)
• JAGS_bridgesampling_posterior unable to add a parameter via add_parameters

Features

• interpret function for creating textual summaries based on inference and samples objects

version 0.1.1

Fixes

• plot_posterior fails with only mu & PET samples (Issue #5)
• ordering by “probabilities” does not work in ‘plot_models’ (Issue #3)
• BF goes to NaN when only a single model is present in ‘models_inference’ (Issue #2)
• summary tables unit tests unable to deal with numerical precision
• problems with aggregating samples across multiple spikes in `plot_posterior’

Features

• allow density.prior with range lower == upper (Issue #4)
• moving rstan towards suggested packages

version 0.1.0

• published on CRAN

version 0.0.0.9010

• plotting functions for models

version 0.0.0.9009

• plotting functions for posterior samples

version 0.0.0.9008

• plotting functions for mixture of priors

version 0.0.0.9007

• improvements to prior plotting functions

version 0.0.0.9006

• ensemble and model summary tables functions

version 0.0.0.9005

• posterior mixing functions

version 0.0.0.9004

• model-averaging functions

version 0.0.0.9003

• JAGS fitting related functions

version 0.0.0.9002

• JAGS bridgesampling related functions

version 0.0.0.9001

• JAGS model building related functions

version 0.0.0.9000

• priors and related methods