- Fit extended-support Beta models via family
xbeta
thanks to Ioannis Kosmidis. (#1698) - Add a
seed
argument toloo_R2
thanks to Marco Colombo. (#1713)
- Avoid the creation of zombie workers when executing
log_lik
in parallel thanks to Aki Vehtari and Noa Kallioinen. For now,log_lik
will use PSOCK clusters if run in parallel even on Unix systems. To avoid potential speed loss for small models,log_lik
will not useoption(mc.cores)
anymore. These changes may be reverted once the underlying causes of this issue have been fixed. (#1658)
- Improve sampling efficiency of
beta_binomial
models. (#1703)
- Support different Gaussian process kernels in
gp
terms. (#234) - Support stratified
cox
models via the new addition termbhaz
. (#1489) - Support futures for parallelization in the
cmdstanr
backend. (#1684) - Add method
loo_epred
thanks to Aki Vehtari. (#1641) - Add priorsense support via
create_priorsense_data.brmsfit
thanks to Noa Kallioinen. (#1354) - Vectorize censored log likelihoods in the Stan code when possible. (#1657)
- Force Stan to activate threading without altering the Stan code
via argument
force
of functionthreading
. (#1549) - Support moment matching
loo
prediction methods. (#1674)
- Fix a bug that led to partially duplicated Stan code in multilevel terms thanks to Henrik Singmann. (#1651)
- Fix problems with parallel executions of post-processing functions sometimes leaving unused R instances behind. Thanks to Andrew Johnson, Aki Vehtari, and Noa Kallioinen. (#1658)
- Fix several minor bugs. (#1648, #1644, #1672, #1642, #1634, #1666, #1664)
- Refactor some of the internal code base to avoid evaluating many data-dependent quantities several times. (#1653)
- Smartly access internal functions when evaluating non-linear formulas. (#1635)
- Improve the documentation in several places.
- Make argument
loo
optional inloo_moment_match
. - Change the output format of
loo_predict
andloo_linpred
to be more consistent with other post-processing functions.
- Add experimental support for the
pathfinder
andlaplace
algorithms in thecmdstanr
backend. (#1591) - Automatically recompute fit criteria previously stored in the model if potentially results-changing arguments are provided to the criterion method.
- Allow to turn off automatic broadcasting of
constant
priors. - Allow for joint likelihood evaluation in
kfold
via argumentjoint
. - Use several Stan built-in functions implemented since version 2.26 to improve the efficiency of multiple model classes. (#1077)
- Change
make_stancode
andmake_standata
to be aliases ofstancode
andstandata
, respectively. Changeget_prior
to be an alias of a new generic methoddefault_prior
. This enable other packages to define newstancode
,standata
anddefault_prior
methods to generate Stan code and data, and extract the default priors, for their own objects building on brms. Thanks to Ven Popov for helping with this. (#1604) - Change the default prior of the
shape
parameter ofnegbinomial
models toinv_gamma(0.4, 0.3)
thanks to Aki Vehtari. (#1614) - No longer automatically canonicalize the Stan code if cmdstanr is used as backend. (#1544)
- Export
read_csv_as_stanfit
thanks to Ven Popov. (#1619) - Make installation of
shinystan
optional. This means that the package has to be loaded, vialibrary(shinystan)
, beforelaunch_shinystan
can be used. (#1595) - Improve parameter class names in the
summary
output. - Show histograms rather than densities in the
plot
method by default. - Deprecate argument
N
in theplot
method in favor of argumentnvariables
. - Remove deprecated argument
exact_loo
in methodkfold
.
- Remove some remaining uses of Stan's old array syntax.
- Fix a bug in formula parsing of missing values terms with interactions thank to Guido Biele. (#1608)
- Ensure compatibility of
combine_models
with moment matching. (#1603) - Ensure compatibility with the latest
splines2
package version. (#1580) - Fix output of
rmulti_normal
thanks to Ven Popov. (#1588) - Prevent memory leaks when executing
kfold
orreloo
in parallel.
- Switch to the new array syntax of Stan. This increases the version requirements of Stan to >= 2.26.
- Apply the
horseshoe
andR2D2
priors globally, that is, for all additive predictor terms specified in the same formula. (#1492) - Use
as.brmsprior
to transform objects into abrmsprior
. (#1491) - Use matrix data as non-linear covariates. (#1488)
- No longer support the
lasso
prior as it is not a good shrinkage prior and incompatible with the newly implemented global shrinkage prior framework. - No longer support multiple deprecated prior options for categorical and multivariate models after around 3 years of deprecation. (#1420)
- Deprecate argument
newdata
ofget_refmodel.brmsfit()
. (#1502) - Disallow binomial models without
trials
argument after several years of deprecation. (#1501)
- Fix a long-standing bug in the post-processing of spline models that could
lead to non-sensible results if predictions were performed on a different
machine than where the model was originally fitted. Old spline models can be
repaired via
restructure
. Special thanks to Simon Wood, Ruben Arslan, Marta Kołczyńska, Patrick Hogan, and Urs Kalbitzer. (#1465) - Fix a bunch of minor issues occurring for rare feature combinations.
- Model unstructured autocorrelation matrices via the
unstr
term thanks to the help of Sebastian Weber. (#1435) - Model ordinal data with an extra category (non-response or similar)
via the
hurdle_cumulative
family thanks to Stephen Wild. (#1448) - Improve user control over model recompilation via argument
recompile
in post-processing methods that require a compiled Stan model. - Extend control over the
point_estimate
feature inprepare_predictions
via the new argumentndraws_point_estimate
. - Add support for the latent projection available in projpred versions >= 2.4.0. (#1451)
- Fix a Stan syntax error in threaded models with
lasso
priors. (#1427) - Fix Stan compilation issues for some of the more special
link functions such as
cauchit
orsoftplus
. - Fix a bug for predictions in projpred, previously requiring more variables
in
newdata
than necessary. (#1457, #1459, #1460)
- Support regression splines with fixed degrees of freedom
specified via
s(..., fx = TRUE)
. - Reuse user-specified control arguments originally passed
to the Stan backend in
update
and related methods. (#1373, #1378) - Allow to retain unused factors levels via
drop_unused_levels = FALSE
inbrm
and related functions. (#1346) - Automatically update old default priors based on new input when
when updating models via
update.brmsfit
. (#1380) - Allow to use
dirichlet
priors for more parameter types. (#1165)
- Improve efficiency of converting models fitted with
backend = "cmdstanr"
tostanfit
objects thanks to Simon Mills and Jacob Socolar. (#1331) - Allow for more
O1
optimization of brms-generated Stan models thanks to Aki Vehtari. (#1382)
- Fix problems with missing boundaries of
sdme
parameters in models with known response standard errors thanks to Solomon Kurz. (#1348) - Fix Stan code of
gamma
models withsoftplus
link. - Allow for more flexible data inputs to
brm_multiple
. (#1383) - Ensure that
control_params
returns the right values for models fitted with thecmdstanr
backend. (#1390) - Fix problems in multivariate spline models when using
the
subset
addition term. (#1385)
- Add full user control for boundaries of most parameters via the
lb
andub
arguments ofset_prior
and related functions. (#878, #1094) - Add family
logistic_normal
for simplex responses. (#1274) - Add argument
future_args
tokfold
andreloo
for additional control over parallel execution via futures. - Add families
beta_binomial
&zero_inflated_beta_binomial
for potentially over-dispersed and zero-inflated binomial response models thanks to Hayden Rabel. (#1319 & #1311) - Display
ppd_*
plots inpp_check
via argumentprefix
. (#1313) - Support the
log
link in binomial and beta type families. (#1316) - Support projpred's augmented-data projection. (#1292, #1294)
- Argument
brms_seed
has been added toget_refmodel.brmsfit()
. (#1287) - Deprecate argument
inits
in favor ofinit
for consistency with the Stan backends. - Improve speed of the
summary
method for high-dimensional models. (#1330)
- Fix Stan code of threaded multivariate models thanks to Anirban Mukherjee. (#1277)
- Fix usage of
int_conditions
inconditional_smooths
thanks to Urs Kalbitzer. (#1280) - Fix an error sometimes occurring for multilevel (reference) models in
projpred
's K-fold CV. (#1286) - Fix response values in
make_standata
forbernoulli
families when only 1s are present thanks to Facundo Munoz. (#1298) - Fix
pp_check
for censored responses to work for all plot types thanks to Hayden Rabel. (#1327) - Ensure that argument
overwrite
inadd_criterion
works as expected for all criteria thanks to Andrew Milne. (#1323) - Fix a problem in
launch_shinystan
occurring when warmup draws were saved thanks to Frank Weber. (#1257, #1329) - Fix numerical stability problems in
log_lik
for ordinal models. (#1192)
- Move
projpred
fromImports:
toSuggests:
. This has the important implication that users need to load or attachprojpred
themselves if they want to use it (the more common case is probably attaching, which is achieved bylibrary(projpred)
). (#1222)
- Ensure that argument
overwrite
inadd_criterion
is working as intended thanks to Ruben Arslan. (#1219) - Fix a bug in
get_refmodel.brmsfit()
(i.e., when usingprojpred
for a"brmsfit"
) causing offsets not to be recognized. (#1220) - Several further minor bug fixes.
- Fix a bug causing problems during post-processing of models
fitted with older versions of brms and the
cmdstanr
backend thanks to Riccardo Fusaroli. (#1218)
- Support several methods of the
posterior
package. (#1204) - Substantially extend compatibility of
brms
models withemmeans
thanks to Mattan S. Ben-Shachar. (#907, #1134) - Combine missing value (
mi
) terms withsubset
addition terms. (#1063) - Expose function
get_dpar
for use in the post-processing of custom families thank to Martin Modrak. (#1131) - Support the
squareplus
link function in all families and distributional parameters that also allow for thelog
link function. - Add argument
incl_thres
toposterior_linpred.brmsfit()
allowing to subtract the threshold-excluding linear predictor from the thresholds in case of an ordinal family. (#1137) - Add a
"mock"
backend option to facilitate testing thanks to Martin Modrak. (#1116) - Add option
file_refit = "always"
to always overwrite models stored via thefile
argument. (#1151) - Initial GPU support via OpenCL thanks to the help Rok Češnovar. (#1166)
- Support argument
robust
in methodhypothesis
. (#1170) - Vectorize the Stan code of custom likelihoods via
argument
loop
ofcustom_family
. (#1084) - Experimentally allow category specific effects for
ordinal
cumulative
models. (#1060) - Regenerate Stan code of an existing model via
argument
regenerate
of methodstancode
. - Support
expose_functions
for models fitted with thecmdstanr
backend thanks to Sebastian Weber. (#1176) - Support
log_prob
and related functionality in models fitted with thecmdstanr
backend via functionadd_rstan_model
. (#1184)
- Remove use of
cbind
to express multivariate models after over two years of deprecation (please usemvbind
instead). - Method
posterior_linpred(transform = TRUE)
is now equal toposterior_epred(dpar = "mu")
and no longer deprecated. - Refactor and extend internal post-processing functions for ordinal and categorical models thanks to Frank Weber. (#1159)
- Ignore
NA
values in interval censored boundaries as long as they are unused. (#1070) - Take offsets into account when deriving default priors for overall intercept parameters. (#923)
- Soft deprecate measurement error (
me
) terms in favor of the more general and consistent missing value (mi
) terms. (#698)
- Fix an issue in the post-processing of non-normal ARMA models thanks to Thomas Buehrens. (#1149)
- Fix an issue with default baseline hazard knots in
cox
models thanks to Malcolm Gillies. (#1143) - Fix a bug in non-linear models caused by accidental merging of operators in the non-linear formula thanks to Fernando Miguez. (#1142)
- Correctly trigger a refit for
file_refit = "on_change"
if factor level names have changed thanks to Martin Modrak. (#1128) - Validate factors in
validate_newdata
even when they are simultaneously used as predictors and grouping variables thanks to Martin Modrak. (#1141) - Fix a bug in the Stan code generation of threaded mixture models with predicted mixture probabilities thanks to Riccardo Fusaroli. (#1150)
- Remove duplicated Stan code related to the
horseshoe
prior thanks to Max Joseph. (#1167) - Fix an issue in the post-processing of non-looped non-linear parameters thanks to Sebastian Weber.
- Fix an issue in the Stan code of threaded non-looped non-linear models thanks to Sebastian Weber. (#1175)
- Fix problems in the post-processing of multivariate meta-analytic models that could lead to incorrect handling of known standard errors.
- Turn off normalization in the Stan model via argument
normalize
. to increase sampling efficiency thanks to Andrew Johnson. (#1017, #1053) - Enable
posterior_predict
for truncated continuous models even if the required CDF or quantile functions are unavailable. - Update and export
validate_prior
to validate priors supplied by the user. - Add support for within-chain threading with
rstan (Stan >= 2.25)
backend. - Apply the R2-D2 shrinkage prior to population-level coefficients
via function
R2D2
to be used inset_prior
. - Extend support for
arma
correlation structures in non-normal families. - Extend scope of variables passed via
data2
for use in the evaluation of most model terms. - Refit models previously stored on disc only when necessary thanks to
Martin Modrak. The behavior can be controlled via
file_refit
. (#1058) - Allow for a finer tuning of informational messages printed in
brm
via thesilent
argument. (#1076) - Allow
stanvars
to alter distributional parameters. (#1061) - Allow
stanvars
to be used inside threaded likelihoods. (#1111)
- Improve numerical stability of ordinal sequential models
(families
sratio
andcratio
) thanks to Andrew Johnson. (#1087)
- Allow fitting
multinomial
models with thecmdstanr
backend thanks to Andrew Johnson. (#1033) - Allow user-defined Stan functions in threaded models. (#1034)
- Allow usage of the
:
operator in autocorrelation terms. - Fix Stan code generation when specifying coefficient-level priors on spline terms.
- Fix numerical issues occurring in edge cases during post-processing of Gaussian processes thanks to Marta Kołczyńska.
- Fix an error during post-processing of new levels in multi-membership terms thanks to Guilherme Mohor.
- Fix a bug in the Stan code of threaded
wiener
drift diffusion models thanks to the GitHub user yanivabir. (#1085) - Fix a bug in the threaded Stan code for GPs with categorical
by
variables thanks to Reece Willoughby. (#1081) - Fix a bug in the threaded Stan code when using QR decomposition thanks to Steve Bronder. (#1086)
- Include offsets in
emmeans
related methods thanks to Russell V. Lenth. (#1096)
- Support
projpred
version 2.0 for variable selection in generalized linear and additive multilevel models thanks to Alejandro Catalina. - Support
by
variables in multi-membership terms. - Use Bayesian bootstrap in
loo_R2
.
- Allow non-linear terms in threaded models.
- Allow multi-membership terms in threaded models.
- Allow
se
addition terms in threaded models. - Allow
categorical
families in threaded models. - Fix updating of parameters in
loo_moment_match
. - Fix facet labels in
conditional_effects
thanks to Isaac Petersen. (#1014)
- Experimentally support within-chain parallelization via
reduce_sum
using argumentthreads
inbrm
thanks to Sebastian Weber. (#892) - Add algorithm
fixed_param
to sample from fixed parameter values. (#973) - No longer remove
NA
values indata
if there are unused because of thesubset
addition argument. (#895) - Combine
by
variables and within-group correlation matrices in group-level terms. (#674) - Add argument
robust
to thesummary
method. (#976) - Parallelize evaluation of the
posterior_predict
andlog_lik
methods via argumentcores
. (#819) - Compute effective number of parameters in
kfold
. - Show prior sources and vectorization in the
print
output ofbrmsprior
objects. (#761) - Store unused variables in the model's data frame via
argument
unused
of functionbrmsformula
. - Support posterior mean predictions in
emmeans
viadpar = "mean"
thanks to Russell V. Lenth. (#993) - Improve control of which parameters should be saved via
function
save_pars
and corresponding argument inbrm
. (#746) - Add method
posterior_smooths
to computing predictions of individual smooth terms. (#738) - Allow to display grouping variables in
conditional_effects
using theeffects
argument. (#1012)
- Improve sampling efficiency for a lot of models by using Stan's GLM-primitives even in non-GLM cases. (#984)
- Improve sampling efficiency of multilevel models with within-group covariances thanks to David Westergaard. (#977)
- Deprecate argument
probs
in theconditional_effects
method in favor of argumentprob
.
- Fix a problem in
pp_check
inducing wronger observation orders in time series models thanks to Fiona Seaton. (#1007) - Fix multiple problems with
loo_moment_match
that prevented it from working for some more complex models.
- Support the Cox proportional hazards model for
time-to-event data via family
cox
. (#230, #962) - Support method
loo_moment_match
, which can be used to update aloo
object when Pareto k estimates are large.
- Improve the prediction behavior in post-processing methods
when sampling new levels of grouping factors via
sample_new_levels = "uncertainty"
. (#956)
- Fix minor problems with MKL on CRAN.
- Fix shape parameters across multiple monotonic terms via argument
id
in functionmo
to ensure conditionally monotonic effects. (#924) - Support package
rtdists
as additional backend ofwiener
distribution functions thanks to the help of Henrik Singmann. (#385)
- Fix generated Stan Code of models with improper global priors and
constant
priors on some coefficients thanks to Frank Weber. (#919) - Fix a bug in
conditional_effects
occurring for categorical models with matrix predictors thanks to Jamie Cranston. (#933)
- Adjust behavior of the
rate
addition term so that it also affects theshape
parameter innegbinomial
models thanks to Edward Abraham. (#915) - Adjust the default inverse-gamma prior on length-scale parameters of Gaussian processes to be less extreme in edge cases thanks to Topi Paananen.
- Constrain ordinal thresholds to sum to zero via argument
threshold
in ordinal family functions thanks to the help of Marta Kołczyńska. - Support
posterior_linpred
as method inconditional_effects
. - Use
std_normal
in the Stan code for improved efficiency. - Add arguments
cor
,id
, andcov
to the functionsgr
andmm
for easy specification of group-level correlation structures. - Improve workflow to feed back brms-created models which were fitted somewhere else back into brms. (#745)
- Improve argument
int_conditions
inconditional_effects
to work for all predictors not just interactions. - Support multiple imputation of data passed via
data2
inbrm_multiple
. (#886) - Fully support the
emmeans
package thanks to the help of Russell V. Lenth. (#418) - Control the within-block position of Stan code added via
stanvar
using theposition
argument.
- Fix issue in Stan code of models with multiple
me
terms thanks to Chris Chatham. (#855, #856) - Fix scaling problems in the estimation of ordinal models with multiple threshold vectors thanks to Marta Kołczyńska and Rok Češnovar.
- Allow usage of
std_normal
inset_prior
thanks to Ben Goodrich. (#867) - Fix Stan code of distributional models with
weibull
,frechet
, orinverse.gaussian
families thanks to Brian Huey and Jack Caster. (#879) - Fix Stan code of models which are truncated and weighted at the same time thanks to Michael Thompson. (#884)
- Fix Stan code of multivariate models with custom families and data variables passed to the likelihood thanks to Raoul Wolf. (#906)
- Reduce minimal scale of several default priors from 10 to 2.5. The resulting priors should remain weakly informative.
- Automatically group observations in
gp
for increased efficiency. - Rename
parse_bf
tobrmsterms
and deprecate the former function. - Rename
extract_draws
toprepare_predictions
and deprecate the former function. - Deprecate using a model-dependent
rescor
default. - Deprecate argument
cov_ranef
inbrm
and related functions. - Improve several internal interfaces. This should not have any user-visible changes.
- Simplify the parameterization of the horseshoe prior thanks to Aki Vehtari. (#873)
- Store fixed distributional parameters as regular draws so that they behave as if they were estimated in post-processing methods.
- Fix parameters to constants via the
prior
argument. (#783) - Specify autocorrelation terms directly in the model formula. (#708)
- Translate integer covariates in non-linear formulas to integer arrays in Stan.
- Estimate
sigma
in combination with fixed correlation matrices via autocorrelation termfcor
. - Use argument
data2
inbrm
and related functions to pass data objects which cannot be passed viadata
. The usage ofdata2
will be extended in future versions. - Compute pointwise log-likelihood values via
log_lik
for non-factorizable Student-t models. (#705)
- Fix output of
posterior_predict
formultinomial
models thanks to Ivan Ukhov. - Fix selection of group-level terms via
re_formula
in multivariate models thanks to Maxime Dahirel. (#834) - Enforce correct ordering of terms in
re_formula
thanks to @ferberkl. (#844) - Fix post-processing of multivariate multilevel models when multiple IDs are used for the same grouping factor thanks to @lott999. (#835)
- Store response category names of ordinal models in the
output of
posterior_predict
again thanks to Mattew Kay. (#838) - Handle
NA
values more consistently inposterior_table
thanks to Anna Hake. (#845) - Fix a bug in the Stan code of models with multiple monotonic varying effects across different groups thanks to Julian Quandt.
- Rename
offset
variables tooffsets
in the generated Stan code as the former will be reserved in the new stanc3 compiler.
- Fix version requirement of the
loo
package. - Fix effective sample size note in the
summary
output. (#824) - Fix an edge case in the handling of covariates in special terms thanks to Andrew Milne. (#823)
- Allow restructuring objects multiple times with different brms versions thanks to Jonathan A. Nations. (#828)
- Fix validation of ordered factors in
newdata
thanks to Andrew Milne. (#830)
- Support grouped ordinal threshold vectors via addition
argument
resp_thres
. (#675) - Support method
loo_subsample
for performing approximate leave-one-out cross-validation for large data. - Allow storing more model fit criteria via
add_criterion
. (#793)
- Fix prediction uncertainties of new group levels for
sample_new_levels = "uncertainty"
thanks to Dominic Magirr. (#779) - Fix problems when using
pp_check
on censored models thanks to Andrew Milne. (#744) - Fix error in the generated Stan code of multivariate
zero_inflated_binomial
models thanks to Raoul Wolf. (#756) - Fix predictions of spline models when using addition
argument
subset
thanks to Ruben Arslan. - Fix out-of-sample predictions of AR models when predicting more than one step ahead.
- Fix problems when using
reloo
orkfold
with CAR models. - Fix problems when using
fitted(..., scale = "linear")
with multinomial models thanks to Santiago Olivella. (#770) - Fix problems in the
as.mcmc
method for thinned models thanks to @hoxo-m. (#811) - Fix problems in parsing covariates of special effects terms thanks to Riccardo Fusaroli (#813)
- Rename
marginal_effects
toconditional_effects
andmarginal_smooths
toconditional_smooths
. (#735) - Rename
stanplot
tomcmc_plot
. - Add method
pp_expect
as an alias offitted
. (#644) - Model fit criteria computed via
add_criterion
are now stored in thebrmsfit$criteria
slot. - Deprecate
resp_cat
in favor ofresp_thres
. - Deprecate specifying global priors on regression coefficients in categorical and multivariate models.
- Improve names of weighting methods in
model_weights
. - Deprecate reserved variable
intercept
in favor ofIntercept
. - Deprecate argument
exact_match
in favor offixed
. - Deprecate functions
add_loo
andadd_waic
in favor ofadd_criterion
.
- Improve convergence diagnostics in the
summary
output. (#712) - Use primitive Stan GLM functions whenever possible. (#703)
- Pass real and integer data vectors to custom families via
the addition arguments
vreal
andvint
. (#707) - Model compound symmetry correlations via
cor_cosy
. (#403) - Predict
sigma
in combination with several autocorrelation structures. (#403) - Use addition term
rate
to conveniently handle denominators of rate responses in log-linear models. - Fit BYM2 CAR models via
cor_car
thanks to the case study and help of Mitzi Morris.
- Substantially improve the sampling efficiency of SAR models thanks to the GitHub user aslez. (#680)
- No longer allow changing the boundaries of autocorrelation parameters.
- Set the number of trials to 1 by default in
marginal_effects
if not specified otherwise. (#718) - Use non-standard evaluation for addition terms.
- Name temporary intercept parameters more consistently in the Stan code.
- Fix problems in the post-processing of
me
terms with grouping factors thanks to the GitHub user tatters. (#706) - Allow grouping variables to start with a dot thanks to Bruno Nicenboim. (#679)
- Allow the
horseshoe
prior in categorical and related models thanks to the Github user tatters. (#678) - Fix extraction of prior samples for overall intercepts in
prior_samples
thanks to Jonas Kristoffer Lindelov. (#696) - Allow underscores to be used in category names of categorical responses thanks to Emmanuel Charpentier. (#672)
- Fix Stan code of multivariate models with multi-membership terms thanks to the Stan discourse user Pia.
- Improve checks for non-standard variable names thanks to Ryan Holbrook. (#721)
- Fix problems when plotting facetted spaghetti plots
via
marginal_smooths
thanks to Gavin Simpson. (#740)
- Specify non-linear ordinal models. (#623)
- Allow to fix thresholds in ordinal mixture models (#626)
- Use the
softplus
link function in various families. (#622) - Use QR decomposition of design matrices via argument
decomp
ofbrmsformula
thanks to the help of Ben Goodrich. (#640) - Define argument
sparse
separately for each model formula. - Allow using
bayes_R2
andloo_R2
with ordinal models. (#639) - Support
cor_arma
in non-normal models. (#648)
- Change the parameterization of monotonic effects to improve their interpretability. (#578)
- No longer support the
cor_arr
andcor_bsts
correlation structures after a year of deprecation. - Refactor internal evaluation of special predictor terms.
- Improve penalty of splines thanks to Ben Goodrich and Ruben Arslan.
- Fix a problem when applying
marginal_effects
to measurement error models thanks to Jonathan A. Nations. (#636) - Fix computation of log-likelihood values for weighted mixture models.
- Fix computation of fitted values for truncated lognormal and weibull models.
- Fix checking of response boundaries for models with missing values thanks to Lucas Deschamps.
- Fix Stan code of multivariate models with both residual correlations and missing value terms thanks to Solomon Kurz.
- Fix problems with interactions of special terms
when extracting variable names in
marginal_effects
. - Allow compiling a model in
brm_multiple
without sampling thanks to Will Petry. (#671)
- Fit multinomial models via family
multinomial
. (#463) - Fit Dirichlet models via family
dirichlet
. (#463) - Fit conditional logistic models using the
categorical
andmultinomial
families together with non-linear formula syntax. (#560) - Choose the reference category of
categorical
and related families via argumentrefcat
of the corresponding family functions. - Use different subsets of the data in different univariate parts
of a multivariate model via addition argument
subset
. (#360) - Control the centering of population-level design matrices
via argument
center
ofbrmsformula
and related functions. - Add an
update
method forbrmsfit_multiple
objects. (#615) - Split folds after
group
in thekfold
method. (#619)
- Deprecate
compare_ic
and instead recommendloo_compare
for the comparison ofloo
objects to ensure consistency between packages. (#414) - Use the glue package in the Stan code generation. (#549)
- Introduce
mvbind
to eventually replacecbind
in the formula syntax of multivariate models. - Validate several sampling-related arguments in
brm
before compiling the Stan model. (#576) - Show evaluated vignettes on CRAN again. (#591)
- Export function
get_y
which is used to extract response values frombrmsfit
objects.
- Fix an error when trying to change argument
re_formula
inbayes_R2
thanks to the GitHub user emieldl. (#592) - Fix occasional problems when running chains in parallel via the future package thanks to Jared Knowles. (#579)
- Ensure correct ordering of response categories in ordinal models thanks to Jonas Kristoffer Lindelov. (#580)
- Ignore argument
resp
ofmarginal_effects
in univariate models thanks to Vassilis Kehayas. (#589) - Correctly disable cell-mean coding in varying effects.
- Allow to fix parameter
ndt
in drift diffusion models. - Fix Stan code for t-distributed varying effects thanks to Ozgur Asar.
- Fix an error in the post-processing of monotonic effects occurring for multivariate models thanks to James Rae. (#598)
- Fix lower bounds in truncated discrete models.
- Fix checks of the original data in
kfold
thanks to the GitHub user gcolitti. (#602) - Fix an error when applying the
VarCorr
method to meta-analytic models thanks to Michael Scharkow. (#616)
- Fit approximate and non-isotropic Gaussian processes via
gp
. (#540) - Enable parallelization of model fitting in
brm_multiple
via the future package. (#364) - Perform posterior predictions based on k-fold cross-validation
via
kfold_predict
. (#468) - Indicate observations for out-of-sample predictions in
ARMA models via argument
oos
ofextract_draws
. (#539)
- Allow factor-like variables in smooth terms. (#562)
- Make plotting of
marginal_effects
more robust to the usage of non-standard variable names. - Deactivate certain data validity checks when using custom families.
- Improve efficiency of adjacent category models.
- No longer print informational messages from the Stan parser.
- Fix an issue that could result in a substantial efficiency drop of various post-processing methods for larger models.
- Fix an issue when that resulted in an error when
using
fitted(..., scale = "linear")
with ordinal models thanks to Andrew Milne. (#557) - Allow setting priors on the overall intercept in sparse models.
- Allow sampling from models with only a single observation that also contain an offset thanks to Antonio Vargas. (#545)
- Fix an error when sampling from priors in mixture models thanks to Jacki Buros Novik. (#542)
- Fix a problem when trying to sample from priors of parameter transformations.
- Allow using
marginal_smooths
with ordinal models thanks to Andrew Milne. (#570) - Fix an error in the post-processing of
me
terms thanks to the GitHub user hlluik. (#571) - Correctly update
warmup
samples when usingupdate.brmsfit
.
- Fit factor smooth interactions thanks to Simon Wood.
- Specify separate priors for thresholds in ordinal models. (#524)
- Pass additional arguments to
rstan::stan_model
via argumentstan_model_args
inbrm
. (#525) - Save model objects via argument
file
inadd_ic
after adding model fit criteria. (#478) - Compute density ratios based on MCMC samples via
density_ratio
. - Ignore offsets in various post-processing methods via
argument
offset
. - Update addition terms in formulas via
update_adterms
.
- Improve internal modularization of smooth terms.
- Reduce size of internal example models.
- Correctly plot splines with factorial covariates via
marginal_smooths
. - Allow sampling from priors in intercept only models thanks to Emmanuel Charpentier. (#529)
- Allow logical operators in non-linear formulas.
- Improve
marginal_effects
to better display ordinal and categorical models via argumentcategorical
. (#491, #497) - Improve method
kfold
to offer more options for specifying omitted subsets. (#510) - Compute estimated values of non-linear parameters via
argument
nlpar
in methodfitted
. - Disable automatic cell-mean coding in model formulas without
an intercept via argument
cmc
ofbrmsformula
and related functions thanks to Marie Beisemann. - Allow using the
bridge_sampler
method even if prior samples are drawn within the model. (#485) - Specify post-processing functions of custom families
directly in
custom_family
. - Select a subset of coefficients in
fixef
,ranef
, andcoef
via argumentpars
. (#520) - Allow to
overwrite
already stored fit indices when usingadd_ic
.
- Ignore argument
resp
when post-processing univariate models thanks to Ruben Arslan. (#488) - Deprecate argument
ordinal
ofmarginal_effects
. (#491) - Deprecate argument
exact_loo
ofkfold
. (#510) - Deprecate usage of
binomial
families without specifyingtrials
. - No longer sample from priors of population-level intercepts when using the default intercept parameterization.
- Correctly sample from LKJ correlation priors thanks to Donald Williams.
- Remove stored fit indices when calling
update
on brmsfit objects thanks to Emmanuel Charpentier. (#490) - Fix problems when predicting a single data point using spline models thanks to Emmanuel Charpentier. (#494)
- Set
Post.Prob = 1
ifEvid.Ratio = Inf
in methodhypothesis
thanks to Andrew Milne. (#509) - Ensure correct handling of argument
file
inbrm_multiple
.
- Define custom variables in all of Stan's program blocks
via function
stanvar
. (#459) - Change the scope of non-linear parameters to be global within univariate models. (#390)
- Allow to automatically group predictor values in Gaussian
processes specified via
gp
. This may lead to a considerable increase in sampling efficiency. (#300) - Compute LOO-adjusted R-squared using method
loo_R2
. - Compute non-linear predictors outside of a loop over
observations by means of argument
loop
inbrmsformula
. - Fit non-linear mixture models. (#456)
- Fit censored or truncated mixture models. (#469)
- Allow
horseshoe
andlasso
priors to be set on special population-level effects. - Allow vectors of length greater one to be passed to
set_prior
. - Conveniently save and load fitted model objects in
brm
via argumentfile
. (#472) - Display posterior probabilities in the output of
hypothesis
.
- Deprecate argument
stan_funs
inbrm
in favor of using thestanvars
argument for the specification of custom Stan functions. - Deprecate arguments
flist
and...
innlf
. - Deprecate argument
dpar
inlf
andnlf
.
- Allow custom families in mixture models thanks to Noam Ross. (#453)
- Ensure compatibility with mice version 3.0. (#455)
- Fix naming of correlation parameters of group-level terms with multiple subgroups thanks to Kristoffer Magnusson. (#457)
- Improve scaling of default priors in
lognormal
models (#460). - Fix multiple problems in the post-processing of categorical models.
- Fix validation of nested grouping factors in post-processing methods when passing new data thanks to Liam Kendall.
- Allow censoring and truncation in zero-inflated and hurdle models. (#430)
- Export zero-inflated and hurdle distribution functions.
- Improve sampling efficiency of the ordinal families
cumulative
,sratio
, andcratio
. (#433) - Allow to specify a single k-fold subset in method
kfold
. (#441)
- Fix a problem in
launch_shinystan
due to which the maximum treedepth was not correctly displayed thanks to Paul Galpern. (#431)
- Extend
cor_car
to support intrinsic CAR models in pairwise difference formulation thanks to the case study of Mitzi Morris. - Compute
loo
and related methods for non-factorizable normal models.
- Rename quantile columns in
posterior_summary
. This affects the output ofpredict
and related methods ifsummary = TRUE
. (#425) - Use hashes to check if models have the same response values when performing model comparisons. (#414)
- No longer set
pointwise
dynamically inloo
and related methods. (#416) - No longer show information criteria in the summary output.
- Simplify internal workflow to implement native response distributions. (#421)
- Allow
cor_car
in multivariate models with residual correlations thanks to Quentin Read. (#427) - Fix a problem in the Stan code generation of distributional
beta
models thanks to Hans van Calster. (#404) - Fix
launch_shinystan.brmsfit
so that all parameters are now shown correctly in the diagnose tab. (#340)
- Specify custom response distributions with function
custom_family
. (#381) - Model missing values and measurement error in responses using the
mi
addition term. (#27, #343) - Allow missing values in predictors using
mi
terms on the right-hand side of model formulas. (#27) - Model interactions between the special predictor terms
mo
,me
, andmi
. (#313) - Introduce methods
model_weights
andloo_model_weights
providing several options to compute model weights. (#268) - Introduce method
posterior_average
to extract posterior samples averaged across models. (#386) - Allow hyperparameters of group-level effects to vary over the levels of a
categorical covariate using argument
by
in functiongr
. (#365) - Allow predictions of measurement-error models with new data. (#335)
- Pass user-defined variables to Stan via
stanvar
. (#219, #357) - Allow ordinal families in mixture models. (#389)
- Model covariates in multi-membership structures that vary over the levels of
the grouping factor via
mmc
terms. (#353) - Fit shifted log-normal models via family
shifted_lognormal
. (#218) - Specify nested non-linear formulas.
- Introduce function
make_conditions
to ease preparation of conditions formarginal_effects
.
- Change the parameterization of
weibull
andexgaussian
models to be consistent with other model classes. Post-processing of related models fitted with earlier version ofbrms
is no longer possible. - Treat integer responses in
ordinal
models as directly indicating categories even if the lowest integer is not one. - Improve output of the
hypothesis
method thanks to the ideas of Matti Vuorre. (#362) - Always plot
by
variables as facets inmarginal_smooths
. - Deprecate the
cor_bsts
correlation structure.
- Allow the
:
operator to combine groups in multi-membership terms thanks to Gang Chen. - Avoid an unexpected error when calling
LOO
with argumentreloo = TRUE
thanks to Peter Konings. (#348) - Fix problems in
predict
when applied to categorical models thanks to Lydia Andreyevna Krasilnikova and Thomas Vladeck. (#336, #345) - Allow truncation in multivariate models with missing values thanks to Malte Lau Petersen. (#380)
- Force time points to be unique within groups in autocorrelation structures thanks to Ruben Arslan. (#363)
- Fix problems when post-processing multiple uncorrelated group-level terms of the same grouping factor thanks to Ivy Jansen. (#374)
- Fix a problem in the Stan code of multivariate
weibull
andfrechet
models thanks to the GitHub user philj1s. (#375) - Fix a rare error when post-processing
binomial
models thanks to the GitHub user SeanH94. (#382) - Keep attributes of variables when preparing the
model.frame
thanks to Daniel Luedecke. (#393)
- Fit models on multiple imputed datasets via
brm_multiple
thanks to Ruben Arslan. (#27) - Combine multiple
brmsfit
objects via functioncombine_models
. - Compute model averaged posterior predictions with method
pp_average
. (#319) - Add new argument
ordinal
tomarginal_effects
to generate special plots for ordinal models thanks to the idea of the GitHub user silberzwiebel. (#190) - Use informative inverse-gamma priors for length-scale parameters of Gaussian processes. (#275)
- Compute hypotheses for all levels of a grouping factor at once using
argument
scope
in methodhypothesis
. (#327) - Vectorize user-defined
Stan
functions exported viaexport_functions
using argumentvectorize
. - Allow predicting new data in models with ARMA autocorrelation structures.
- Correctly recover noise-free coefficients through
me
terms thanks to Ruben Arslan. As a side effect, it is no longer possible to define priors on noise-freeXme
variables directly, but only on their hyper-parametersmeanme
andsdme
. - Fix problems in renaming parameters of the
cor_bsts
structure thanks to Joshua Edward Morten. (#312) - Fix some unexpected errors when predicting from ordinal models thanks to David Hervas and Florian Bader. (#306, #307, #331)
- Fix problems when estimating and predicting multivariate ordinal models thanks to David West. (#314)
- Fix various minor problems in autocorrelation structures thanks to David West. (#320)
- Export the helper functions
posterior_summary
andposterior_table
both being used to summarize posterior samples and predictions.
- Fix incorrect computation of intercepts in
acat
andcratio
models thanks to Peter Phalen. (#302) - Fix
pointwise
computation ofLOO
andWAIC
in multivariate models with estimated residual correlation structure. - Fix problems in various S3 methods sometimes requiring unused variables to be
specified in
newdata
. - Fix naming of Stan models thanks to Hao Ran Lai.
This is the second major release of brms
. The main new feature are generalized
multivariate models, which now support everything already possible in univariate
models, but with multiple response variables. Further, the internal structure of
the package has been improved considerably to be easier to maintain and extend
in the future. In addition, most deprecated functionality and arguments have
been removed to provide a clean new start for the package. Models fitted with
brms
1.0 or higher should remain fully compatible with brms
2.0.
- Add support for generalized multivariate models, where each of the univariate
models may have a different family and autocorrelation structure. Residual
correlations can be estimated for multivariate
gaussian
andstudent
models. All features supported in univariate models are now also available in multivariate models. (#3) - Specify different formulas for different categories in
categorical
models. - Add weakly informative default priors for the parameter class
Intercept
to improve convergence of more complex distributional models. - Optionally display the MC standard error in the
summary
output. (#280) - Add argument
re.form
as an alias ofre_formula
to the methodsposterior_predict
,posterior_linpred
, andpredictive_error
for consistency with other packages making use of these methods. (#283)
- Refactor many parts of the package to make it more consistent and easier to extend.
- Show the link functions of all distributional parameters in the
summary
output. (#277) - Reduce working memory requirements when extracting posterior samples for use
in
predict
and related methods thanks to Fanyi Zhang. (#224) - Remove deprecated aliases of functions and arguments from the package. (#278)
- No longer support certain prior specifications, which were previously labeled as deprecated.
- Remove the deprecated addition term
disp
from the package. - Remove old versions of methods
fixef
,ranef
,coef
, andVarCorr
. - No longer support models fitted with
brms
< 1.0, which used the multivariate'trait'
syntax originally deprecated inbrms
1.0. - Make posterior sample extraction in the
summary
method cleaner and less error prone. - No longer fix the seed for random number generation in
brm
to avoid unexpected behavior in simulation studies.
- Store
stan_funs
inbrmsfit
objects to allow usingupdate
on models with user-defined Stan functions thanks to Tom Wallis. (#288) - Fix problems in various post-processing methods when applied to models with
the reserved variable
intercept
in group-level terms thanks to the GitHub user ASKurz. (#279) - Fix an unexpected error in
predict
and related methods when settingsample_new_levels = "gaussian"
in models with only one group-level effect. Thanks to Timothy Mastny. (#286)
- Allow setting priors on noise-free variables specified via function
me
. - Add arguments
Ksub
,exact_loo
andgroup
to methodkfold
for defining omitted subsets according to a grouping variable or factor. - Allow addition argument
se
inskew_normal
models.
- Ensure correct behavior of horseshoe and lasso priors in multivariate models thanks to Donald Williams.
- Allow using
identity
links on all parameters of thewiener
family thanks to Henrik Singmann. (#276) - Use reasonable dimnames in the output of
fitted
when returning linear predictors of ordinal models thanks to the GitHub user atrolle. (#274) - Fix problems in
marginal_smooths
occurring for multi-membership models thanks to Hans Tierens.
- Rebuild monotonic effects from scratch to allow specifying interactions with other variables. (#239)
- Introduce methods
posterior_linpred
andposterior_interval
for consistency with other model fitting packages based onStan
. - Introduce function
theme_black
providing a blackggplot2
theme. - Specify special group-level effects within the same terms as ordinary group-level effects.
- Add argument
prob
tosummary
, which allows to control the width of the computed uncertainty intervals. (#259) - Add argument
newdata
to thekfold
method. - Add several arguments to the
plot
method ofmarginal_effects
to improve control over the appearences of the plots.
- Use the same noise-free variables for all model parts in measurement error models. (#257)
- Make names of local-level terms used in the
cor_bsts
structure more informative. - Store the
autocor
argument withinbrmsformula
objects. - Store posterior and prior samples in separate slots in the output of method
hypothesis
. - No longer change the default theme of
ggplot2
when attachingbrms
. (#256) - Make sure signs of estimates are not dropped when rounding to zero in
summary.brmsfit
. (#263) - Refactor parts of
extract_draws
andlinear_predictor
to be more consistent with the rest of the package.
- Do not silence the
Stan
parser when callingbrm
to get informative error messages about invalid priors. - Fix problems with spaces in priors passed to
set_prior
. - Handle non
data.frame
objects correctly inhypothesis.default
. - Fix a problem relating to the colour of points displayed in
marginal_effects
.
- Perform model comparisons based on marginal likelihoods using the methods
bridge_sampler
,bayes_factor
, andpost_prob
all powered by thebridgesampling
package. - Compute a Bayesian version of R-squared with the
bayes_R2
method. - Specify non-linear models for all distributional parameters.
- Combine multiple model formulas using the
+
operator and the helper functionslf
,nlf
, andset_nl
. - Combine multiple priors using the
+
operator. - Split the
nlpar
argument ofset_prior
into the three argumentsresp
,dpar
, andnlpar
to allow for more flexible prior specifications.
- Refactor parts of the package to prepare for the implementation of more flexible multivariate models in future updates.
- Keep all constants in the log-posterior in order for
bridge_sampler
to be working correctly. - Reduce the amount of renaming done within the
stanfit
object. - Rename argument
auxpar
offitted.brmsfit
todpar
. - Use the
launch_shinystan
generic provided by theshinystan
package. - Set
bayesplot::theme_default()
as the defaultggplot2
theme when attachingbrms
. - Include citations of the
brms
overview paper as published in the Journal of Statistical Software.
- Fix problems when calling
fitted
withhurdle_lognormal
models thanks to Meghna Krishnadas. - Fix problems when predicting
sigma
inasym_laplace
models thanks to Anna Josefine Sorensen.
- Fit conditional autoregressive (CAR) models via function
cor_car
thanks to the case study of Max Joseph. - Fit spatial autoregressive (SAR) models via function
cor_sar
. Currently works for familiesgaussian
andstudent
. - Implement skew normal models via family
skew_normal
. Thanks to Stephen Martin for suggestions on the parameterization. - Add method
reloo
to perform exact cross-validation for problematic observations andkfold
to perform k-fold cross-validation thanks to the Stan Team. - Regularize non-zero coefficients in the
horseshoe
prior thanks to Juho Piironen and Aki Vehtari. - Add argument
new_objects
to various post-processing methods to allow for passing of data objects, which cannot be passed vianewdata
. - Improve parallel execution flexibility via the
future
package.
- Improve efficiency and stability of ARMA models.
- Throw an error when the intercept is removed in an ordinal model instead of silently adding it back again.
- Deprecate argument
threshold
inbrm
and instead recommend passingthreshold
directly to the ordinal family functions. - Throw an error instead of a message when invalid priors are passed.
- Change the default value of the
autocor
slot inbrmsfit
objects to an emptycor_brms
object. - Shorten
Stan
code by combining declarations and definitions where possible.
- Fix problems in
pp_check
when the variable specified in argumentx
has attributes thanks to Paul Galpern. - Fix problems when computing fitted values for truncated discrete models based on new data thanks to Nathan Doogan.
- Fix unexpected errors when passing models, which did not properly initialize, to various post-processing methods.
- Do not accidently drop the second dimension of matrices in
summary.brmsfit
for models with only a single observation.
- Fit latent Gaussian processes of one or more covariates via function
gp
specified in the model formula (#221). - Rework methods
fixef
,ranef
,coef
, andVarCorr
to be more flexible and consistent with other post-processing methods (#200). - Generalize method
hypothesis
to be applicable on all objects coercible to adata.frame
(#198). - Visualize predictions via spaghetti plots using argument
spaghetti
inmarginal_effects
andmarginal_smooths
. - Introduce method
add_ic
to store and reuse information criteria in fitted model objects (#220). - Allow for negative weights in multi-membership grouping structures.
- Introduce an
as.array
method forbrmsfit
objects.
- Show output of \R code in HTML vignettes thanks to Ben Goodrich (#158).
- Resolve citations in PDF vignettes thanks to Thomas Kluth (#223).
- Improve sampling efficiency for
exgaussian
models thanks to Alex Forrence (#222). - Also transform data points when using argument
transform
inmarginal_effects
thanks to Markus Gesmann.
- Fix an unexpected error in
marginal_effects
occurring for some models with autocorrelation terms thanks to Markus Gesmann. - Fix multiple problems occurring for models with the
cor_bsts
structure thanks to Andrew Ellis.
- Implement zero-one-inflated beta models via family
zero_one_inflated_beta
. - Allow for more link functions in zero-inflated and hurdle models.
- Ensure full compatibility with
bayesplot
version 1.2.0. - Deprecate addition argument
disp
.
- Fix problems when setting priors on coefficients of auxiliary parameters when also setting priors on the corresponding coefficients of the mean parameter. Thanks to Matti Vuorre for reporting this bug.
- Allow ordered factors to be used as grouping variables thanks to the GitHub user itissid.
- Fit finite mixture models using family function
mixture
. - Introduce method
pp_mixture
to compute posterior probabilities of mixture component memberships thanks to a discussion with Stephen Martin. - Implement different ways to sample new levels of grouping factors in
predict
and related methods through argumentsample_new_levels
. Thanks to Tom Wallis and Jonah Gabry for a detailed discussion about this feature. - Add methods
loo_predict
,loo_linpred
, andloo_predictive_interval
for computing LOO predictions thanks to Aki Vehtari and Jonah Gabry. - Allow using
offset
in formulas of non-linear and auxiliary parameters. - Allow sparse matrix multiplication in non-linear and distributional models.
- Allow using the
identity
link for all auxiliary parameters. - Introduce argument
negative_rt
inpredict
andposterior_predict
to distinguish responses on the upper and lower boundary inwiener
diffusion models thanks to Guido Biele. - Introduce method
control_params
to conveniently extract control parameters of the NUTS sampler. - Introduce argument
int_conditions
inmarginal_effects
for enhanced plotting of two-way interactions thanks to a discussion with Thomas Kluth. - Improve flexibility of the
conditions
argument ofmarginal_effects
. - Extend method
stanplot
to correctly handle some newmcmc_
plots of thebayesplot
package.
- Improve the
update
method to only recompile models when theStan
code changes. - Warn about divergent transitions when calling
summary
orprint
onbrmsfit
objects. - Warn about unused variables in argument
conditions
when callingmarginal_effects
. - Export and document several distribution functions that were previously kept internal.
- Fix problems with the inclusion of offsets occurring for more complicated formulas thanks to Christian Stock.
- Fix a bug that led to invalid Stan code when sampling from priors in intercept only models thanks to Tom Wallis.
- Correctly check for category specific group-level effects in non-ordinal models thanks to Wayne Folta.
- Fix problems in
pp_check
when specifying argumentnewdata
together with argumentsx
orgroup
. - Rename the last column in the output of
hypothesis
to"star"
in order to avoid problems with zero length column names thanks to the GitHub user puterleat. - Add a missing new line statement at the end of the
summary
output thanks to Thomas Kluth.
- Allow
horseshoe
andlasso
priors to be applied on population-level effects of non-linear and auxiliary parameters. - Force recompiling
Stan
models inupdate.brmsfit
via argumentrecompile
.
- Avoid indexing of matrices in non-linear models to slightly improve sampling speed.
- Fix a severe problem (introduced in version 1.5.0), when predicting
Beta
models thanks to Vivian Lam. - Fix problems when summarizing some models fitted with older version of
brms
thanks to Vivian Lam. - Fix checks of argument
group
in methodpp_check
thanks to Thomas K. - Get arguments
subset
andnsamples
working correctly inmarginal_smooths
.
- Implement the generalized extreme value distribution via family
gen_extreme_value
. - Improve flexibility of the
horseshoe
prior thanks to Juho Piironen. - Introduce auxiliary parameter
mu
as an alternative to specifying effects within theformula
argument in functionbrmsformula
. - Return fitted values of auxiliary parameters via argument
auxpar
of methodfitted
. - Add vignette
"brms_multilevel"
, in which the advanced formula syntax ofbrms
is explained in detail using several examples.
- Refactor various parts of the package to ease implementation of mixture and multivariate models in future updates. This should not have any user visible effects.
- Save the version number of
rstan
in elementversion
ofbrmsfit
objects.
- Fix a rare error when predicting
von_mises
models thanks to John Kirwan.
- Fit quantile regression models via family
asym_laplace
(asymmetric Laplace distribution). - Specify non-linear models in a (hopefully) more intuitive way using
brmsformula
. - Fix auxiliary parameters to certain values through
brmsformula
. - Allow
family
to be specified inbrmsformula
. - Introduce family
frechet
for modelling strictly positive responses. - Allow truncation and censoring at the same time.
- Introduce function
prior_
allowing to specify priors using one-sided formulas orquote
. - Pass priors to
Stan
directly without performing any checks by settingcheck = FALSE
inset_prior
. - Introduce method
nsamples
to extract the number of posterior samples. - Export the main formula parsing function
parse_bf
. - Add more options to customize two-dimensional surface plots created by
marginal_effects
ormarginal_smooths
.
- Change structure of
brmsformula
objects to be more reliable and easier to extend. - Make sure that parameter
nu
never falls below1
to reduce convergence problems when using familystudent
. - Deprecate argument
nonlinear
. - Deprecate family
geometric
. - Rename
cov_fixed
tocor_fixed
. - Make handling of addition terms more transparent by exporting and documenting related functions.
- Refactor helper functions of the
fitted
method to be easier to extend in the future. - Remove many units tests of internal functions and add tests of user-facing functions instead.
- Import some generics from
nlme
instead oflme4
to remove dependency on the latter one. - Do not apply
structure
toNULL
anymore to get rid of warnings in R-devel.
- Fix problems when fitting smoothing terms with factors as
by
variables thanks to Milani Chaloupka. - Fix a bug that could cause some monotonic effects to be ignored in the
Stan
code thanks to the GitHub user bschneider. - Make sure that the data of models with only a single observation are
compatible with the generated
Stan
code. - Handle argument
algorithm
correctly inupdate.brmsfit
. - Fix a bug sometimes causing an error in
marginal_effects
when using familywiener
thanks to Andrew Ellis. - Fix problems in
fitted
when applied tozero_inflated_beta
models thanks to Milani Chaloupka. - Fix minor problems related to the prediction of autocorrelated models.
- Fix a few minor bugs related to the backwards compatibility of multivariate
and related models fitted with
brms
< 1.0.0.
- Introduce the auxiliary parameter
disc
('discrimination') to be used in ordinal models. By default it is not estimated but fixed to one. - Create
marginal_effects
plots of two-way interactions of variables that were not explicitely modeled as interacting.
- Move
rstan
to 'Imports' andRcpp
to 'Depends' in order to avoid loadingrstan
into the global environment automatically.
- Fix a bug leading to unexpected errors in some S3 methods when applied to ordinal models.
- Fit error-in-variables models using function
me
in the model formulae. - Fit multi-membership models using function
mm
in grouping terms. - Add families
exgaussian
(exponentially modified Gaussian distribution) andwiener
(Wiener diffusion model distribution) specifically suited to handle for response times. - Add the
lasso
prior as an alternative to thehorseshoe
prior for sparse models. - Add the methods
log_posterior
,nuts_params
,rhat
, andneff_ratio
forbrmsfit
objects to conveniently access quantities used to diagnose sampling behavior. - Combine chains in method
as.mcmc
using argumentcombine_chains
. - Estimate the auxiliary parameter
sigma
in models with known standard errors of the response by setting argumentsigma
toTRUE
in addition functionse
. - Allow visualizing two-dimensional smooths with the
marginal_smooths
method.
- Require argument
data
to be explicitely specified in all user facing functions. - Refactor the
stanplot
method to usebayesplot
on the backend. - Use the
bayesplot
theme as the default in all plotting functions. - Add the abbreviations
mo
andcs
to specify monotonic and category specific effects respectively. - Rename generated variables in the data.frames returned by
marginal_effects
to avoid potential naming conflicts. - Deprecate argument
cluster
and use the nativecores
argument ofrstan
instead. - Remove argument
cluster_type
as it is no longer required to apply forking. - Remove the deprecated
partial
argument.
- Add the new family
hurdle_lognormal
specifically suited for zero-inflated continuous responses. - Introduce the
pp_check
method to perform various posterior predictive checks using thebayesplot
package. - Introduce the
marginal_smooths
method to better visualize smooth terms. - Allow varying the scale of global shrinkage parameter of the
horseshoe
prior. - Add functions
prior
andprior_string
as aliases ofset_prior
, the former allowing to pass arguments without quotes""
using non-standard evaluation. - Introduce four new vignettes explaining how to fit non-linear models, distributional models, phylogenetic models, and monotonic effects respectively.
- Extend the
coef
method to better handle category specific group-level effects. - Introduce the
prior_summary
method forbrmsfit
objects to obtain a summary of prior distributions applied. - Sample from the prior of the original population-level intercept when
sample_prior = TRUE
even in models with an internal temporary intercept used to improve sampling efficiency. - Introduce methods
posterior_predict
,predictive_error
andlog_lik
as (partial) aliases ofpredict
,residuals
, andlogLik
respectively.
- Improve computation of Bayes factors in the
hypothesis
method to be less influenced by MCMC error. - Improve documentation of default priors.
- Refactor internal structure of some formula and prior evaluating functions. This should not have any user visible effects.
- Use the
bayesplot
package as the new backend ofplot.brmsfit
.
- Better mimic
mgcv
when parsing smooth terms to make sure all arguments are correctly handled. - Avoid an error occurring during the prediction of new data when grouping factors with only a single factor level were supplied thanks to Tom Wallis.
- Fix
marginal_effects
to consistently produce plots for all covariates in non-linear models thanks to David Auty. - Improve the
update
method to better recognize situations where recompliation of theStan
code is necessary thanks to Raphael P.H. - Allow to correctly
update
thesample_prior
argument to value"only"
. - Fix an unexpected error occurring in many S3 methods when the thinning rate is not a divisor of the total number of posterior samples thanks to Paul Zerr.
- Estimate monotonic group-level effects.
- Estimate category specific group-level effects.
- Allow
t2
smooth terms based on multiple covariates. - Estimate interval censored data via the addition argument
cens
in the model formula. - Allow to compute
residuals
also based on predicted values instead of fitted values.
- Use the prefix
bcs
in parameter names of category specific effects and the prefixbm
in parameter names of monotonic effects (instead of the prefixb
) to simplify their identification. - Ensure full compatibility with
ggplot2
version 2.2.
- Fix a bug that could result in incorrect threshold estimates for
cumulative
andsratio
models thanks to Peter Congdon. - Fix a bug that sometimes kept distributional
gamma
models from being compiled thanks to Tim Beechey. - Fix a bug causing an error in
predict
and related methods when two-level factors or logical variables were used as covariates in non-linear models thanks to Martin Schmettow. - Fix a bug causing an error when passing lists to additional arguments of smoothing functions thanks to Wayne Folta.
- Fix a bug causing an error in the
prior_samples
method for models with multiple group-level terms that refer to the same grouping factor thanks to Marco Tullio Liuzza. - Fix a bug sometimes causing an error when calling
marginal_effects
for weighted models.
\subsection{MINOR CHANGES
- Center design matrices inside the Stan code instead of inside
make_standata
. - Get rid of several warning messages occurring on CRAN.
This is one of the largest updates of brms
since its initial release. In
addition to many new features, the multivariate 'trait'
syntax has been
removed from the package as it was confusing for users, required much special
case coding, and was hard to maintain. See help(brmsformula)
for details of
the formula syntax applied in brms
.
- Allow estimating correlations between group-level effects defined across
multiple formulae (e.g., in non-linear models) by specifying IDs in each
grouping term via an extended
lme4
syntax. - Implement distributional regression models allowing to fully predict auxiliary parameters of the response distribution. Among many other possibilities, this can be used to model heterogeneity of variances.
- Zero-inflated and hurdle models do not use multivariate syntax anymore but
instead have special auxiliary parameters named
zi
andhu
defining zero-inflation / hurdle probabilities. - Implement the
von_mises
family to model circular responses. - Introduce the
brmsfamily
function for convenient specification offamily
objects. - Allow predictions of
t2
smoothing terms for new data. - Feature vectors as arguments for the addition argument
trunc
in order to model varying truncation points.
- Remove the
cauchy
family after several months of deprecation. - Make sure that group-level parameter names are unambiguous by adding double underscores thanks to the idea of the GitHub user schmettow.
- The
predict
method now returns predicted probabilities instead of absolute frequencies of samples for ordinal and categorical models. - Compute the linear predictor in the model block of the Stan program instead of in the transformed parameters block. This avoids saving samples of unnecessary parameters to disk. Thanks goes to Rick Arrano for pointing me to this issue.
- Colour points in
marginal_effects
plots if sensible. - Set the default of the
robust
argument toTRUE
inmarginal_effects.brmsfit
.
- Fix a bug that could occur when predicting factorial response variables for new data. Only affects categorical and ordinal models.
- Fix a bug that could lead to duplicated variable names in the Stan code when sampling from priors in non-linear models thanks to Tom Wallis.
- Fix problems when trying to pointwise evaluate non-linear formulae in
logLik.brmsfit
thanks to Tom Wallis. - Ensure full compatibility of the
ranef
andcoef
methods with non-linear models. - Fix problems that occasionally occurred when handling
dplyr
datasets thanks to the GitHub user Atan1988.
- Add support for generalized additive mixed models (GAMMs). Smoothing terms can
be specified using the
s
andt2
functions in the model formula. - Introduce
as.data.frame
andas.matrix
methods forbrmsfit
objects.
- The
gaussian("log")
family no longer implies a log-normal distribution, but a normal distribution with log-link to match the behavior ofglm
. The log-normal distribution can now be specified via familylognormal
. - Update syntax of
Stan
models to match the recommended syntax ofStan
2.10.
- The
ngrps
method should now always return the correct result for non-linear models. - Fix problems in
marginal_effects
for models using the reserved variableintercept
thanks to Frederik Aust. - Fix a bug in the
print
method ofbrmshypothesis
objects that could lead to duplicated and thus invalid row names. - Residual standard deviation parameters of multivariate models are again
correctly displayed in the output of the
summary
method. - Fix problems when using variational Bayes algorithms with
brms
while havingrstan
>= 2.10.0 installed thanks to the GitHub user cwerner87.
- Allow the '/' symbol in group-level terms in the
formula
argument to indicate nested grouping structures. - Allow to compute
WAIC
andLOO
based on the pointwise log-likelihood using argumentpointwise
to substantially reduce memory requirements.
- Add horizontal lines to the errorbars in
marginal_effects
plots for factors.
- Fix a bug that could lead to a cryptic error message when changing some parts
of the model
formula
using theupdate
method. - Fix a bug that could lead to an error when calling
marginal_effects
for predictors that were generated with thebase::scale
function thanks to Tom Wallis. - Allow interactions of numeric and categorical predictors in
marginal_effects
to be passed to theeffects
argument in any order. - Fix a bug that could lead to incorrect results of
predict
and related methods when called withnewdata
in models using thepoly
function thanks to Brock Ferguson. - Make sure that user-specified factor contrasts are always applied in multivariate models.
- Add support for
monotonic
effects allowing to use ordinal predictors without assuming their categories to be equidistant. - Apply multivariate formula syntax in categorical models to considerably increase modeling flexibility.
- Add the addition argument
disp
to define multiplicative factors on dispersion parameters. For linear models,disp
applies to the residual standard deviationsigma
so that it can be used to weight observations. - Treat the fixed effects design matrix as sparse by using the
sparse
argument ofbrm
. This can considerably reduce working memory requirements if the predictors contain many zeros. - Add the
cor_fixed
correlation structure to allow for fixed user-defined covariance matrices of the response variable. - Allow to pass self-defined
Stan
functions via argumentstan_funs
ofbrm
. - Add the
expose_functions
method allowing to expose self-definedStan
functions inR
. - Extend the functionality of the
update
method to allow all model parts to be updated. - Center the fixed effects design matrix also in multivariate models. This may lead to increased sampling speed in models with many predictors.
- Refactor
Stan
code and data generating functions to be more consistent and easier to extent. - Improve checks of user-define prior specifications.
- Warn about models that have not converged.
- Make sure that regression curves computed by the
marginal_effects
method are always smooth. - Allow to define category specific effects in ordinal models directly within
the
formula
argument.
- Fix problems in the generated
Stan
code when using very long non-linear model formulas thanks to Emmanuel Charpentier. - Fix a bug that prohibited to change priors on single standard deviation parameters in non-linear models thanks to Emmanuel Charpentier.
- Fix a bug that prohibited to use nested grouping factors in non-linear models thanks to Tom Wallis.
- Fix a bug in the linear predictor computation within
R
, occurring for ordinal models with multiple category specific effects. This could lead to incorrect outputs ofpredict
,fitted
, andlogLik
for these models. - Make sure that the global
"contrasts"
option is not used when post-processing a model.
- Implement generalized non-linear models, which can be specified with the help
of the
nonlinear
argument inbrm
. - Compute and plot marginal effects using the
marginal_effects
method thanks to the help of Ruben Arslan. - Implement zero-inflated beta models through family
zero_inflated_beta
thanks to the idea of Ali Roshan Ghias. - Allow to restrict domain of fixed effects and autocorrelation parameters using
new arguments
lb
andub
in functionset_prior
thanks to the idea of Joel Gombin. - Add an
as.mcmc
method for compatibility with thecoda
package. - Allow to call the
WAIC
,LOO
, andlogLik
methods with new data.
- Make sure that
brms
is fully compatible withloo
version 0.1.5. - Optionally define the intercept as an ordinary fixed effect to avoid the reparametrization via centering of the fixed effects design matrix.
- Do not compute the WAIC in
summary
by default anymore to reduce computation time of the method for larger models. - The
cauchy
family is now deprecated and will be removed soon as it often has convergence issues and not much practical application anyway. - Change the default settings of the number of chains and warmup samples to the
defaults of
rstan
(i.e.,chains = 4
andwarmup = iter / 2
). - Do not remove bad behaving chains anymore as they may point to general convergence problems that are dangerous to ignore.
- Improve flexibility of the
theme
argument in all plotting functions. - Only show the legend once per page, when computing trace and density plots
with the
plot
method. - Move code of self-defined
Stan
functions toinst/chunks
and incorporate them into the models usingrstan::stanc_builder
. Also, add unit tests for these functions.
- Fix problems when predicting with
newdata
for zero-inflated and hurdle models thanks to Ruben Arslan. - Fix problems when predicting with
newdata
if it is a subset of the data stored in abrmsfit
object thanks to Ruben Arslan. - Fix data preparation for multivariate models if some responses are
NA
thanks to Raphael Royaute. - Fix a bug in the
predict
method occurring for some multivariate models so that it now always returns the predictions of all response variables, not just the first one. - Fix a bug in the log-likelihood computation of
hurdle_poisson
andhurdle_negbinomial
models. This may lead to minor changes in the values obtained byWAIC
andLOO
for these models. - Fix some backwards compatibility issues of models fitted with version <= 0.5.0 thanks to Ulf Koether.
- Use variational inference algorithms as alternative to the NUTS sampler by
specifying argument
algorithm
in thebrm
function. - Implement beta regression models through family
Beta
. - Implement zero-inflated binomial models through family
zero_inflated_binomial
. - Implement multiplicative effects for family
bernoulli
to fit (among others) 2PL IRT models. - Generalize the
formula
argument for zero-inflated and hurdle models so that predictors can be included in only one of the two model parts thanks to the idea of Wade Blanchard. - Combine fixed and random effects estimates using the new
coef
method. - Call the
residuals
method withnewdata
thanks to the idea of Friederike Holz-Ebeling. - Allow new levels of random effects grouping factors in the
predict
,fitted
, andresiduals
methods using argumentallow_new_levels
. - Selectively exclude random effects in the
predict
,fitted
, andresiduals
methods using argumentre_formula
. - Add a
plot
method for objects returned by methodhypothesis
to visualize prior and posterior distributions of the hypotheses being tested.
- Improve evaluation of the response part of the
formula
argument to reliably allow terms with more than one variable (e.g.,y/x ~ 1
). - Improve sampling efficiency of models containing many fixed effects through centering the fixed effects design matrix thanks to Wayne Folta.
- Improve sampling efficiency of models containing uncorrelated random effects
specified by means of
(random || group)
terms informula
thanks to Ali Roshan Ghias. - Utilize user-defined functions in the
Stan
code of ordinal models to improve readability as well as sampling efficiency. - Make sure that model comparisons using
LOO
orWAIC
are only performed when models are based on the same responses. - Use some generic functions of the
lme4
package to avoid unnecessary function masking. This leads to a change in the argument order of methodVarCorr
. - Change the
ggplot
theme in theplot
method through argumenttheme
. - Remove the
n.
prefix in argumentsn.iter
,n.warmup
,n.thin
,n.chains
, andn.cluster
of thebrm
function. The old argument names remain usable as deprecated aliases. - Amend names of random effects parameters to simplify matching with their respective grouping factor levels.
- Fix a bug in the
hypothesis
method that could cause valid model parameters to be falsely reported as invalid. - Fix a bug in the
prior_samples
method that could cause prior samples of parameters of the same class to be artificially correlated. - Fix
Stan
code of linear models with moving-average effects and non-identity link functions so that they no longer contain code related solely to autoregressive effects. - Fix a bug in the evaluation of
formula
that could cause complicated random effects terms to be falsely treated as fixed effects. - Fix several bugs when calling the
fitted
andpredict
methods withnewdata
thanks to Ali Roshan Ghias.
- Add support for zero-inflated and hurdle models thanks to the idea of Scott Baldwin.
- Implement inverse gaussian models through family
inverse.gaussian
. - Allow to specify truncation boundaries of the response variable thanks to the idea of Maciej Beresewicz.
- Add support for autoregressive (AR) effects of residuals, which can be modeled
using the
cor_ar
andcor_arma
functions. - Stationary autoregressive-moving-average (ARMA) effects of order one can now also be fitted using special covariance matrices.
- Implement multivariate student-t models.
- Binomial and ordinal families now support the
cauchit
link function. - Allow family functions to be used in the
family
argument. - Easy access to various
rstan
plotting functions using thestanplot
method. - Implement horseshoe priors to model sparsity in fixed effects coefficients thanks to the idea of Josh Chang.
- Automatically scale default standard deviation priors so that they remain only weakly informative independent on the response scale.
- Report model weights computed by the
loo
package when comparing multiple fitted models.
- Separate the fixed effects Intercept from other fixed effects in the
Stan
code to slightly improve sampling efficiency. - Move autoregressive (AR) effects of the response from the
cor_ar
to thecor_arr
function as the result of implementing AR effects of residuals. - Improve checks on argument
newdata
used in thefitted
andpredict
method. - Method
standata
is now the only way to extract data that was passed toStan
from abrmsfit
object. - Slightly improve
Stan
code for models containing no random effects. - Change the default prior of the degrees of freedom of the
student
family togamma(2,0.1)
. - Improve readability of the output of method
VarCorr
. - Export the
make_stancode
function to give users direct access toStan
code generated bybrms
. - Rename the
brmdata
function tomake_standata
. The former remains usable as a deprecated alias. - Improve documentation to better explain differences in autoregressive effects across R packages.
- Fix a bug that could cause an unexpected error when the
predict
method was called withnewdata
. - Avoid side effects of the
rstan
compilation routines that could occasionally cause R to crash. - Make
brms
work correctly withloo
version 0.1.3 thanks to Mauricio Garnier Villarreal and Jonah Gabry. - Fix a bug that could cause WAIC and LOO estimates to be slightly incorrect for
gaussian
models withlog
link.
- Compute the Watanabe-Akaike information criterion (WAIC) and leave-one-out
cross-validation (LOO) using the
loo
package. - Provide an interface to
shinystan
with S3 methodlaunch_shiny
. - New functions
get_prior
andset_prior
to make prior specifications easier. - Log-likelihood values and posterior predictive samples can now be calculated within R after the model has been fitted.
- Make predictions based on new data using S3 method
predict
. - Allow for customized covariance structures of grouping factors with multiple random effects.
- New S3 methods
fitted
andresiduals
to compute fitted values and residuals, respectively.
- Arguments
WAIC
andpredict
are removed from thebrm
function, as they are no longer necessary. - New argument
cluster_type
in functionbrm
allowing to choose the cluster type created by the parallel package. - Remove chains that fail to initialize while sampling in parallel leaving the other chains untouched.
- Redesign trace and density plots to be faster and more stable.
- S3 method
VarCorr
now always returns covariance matrices regardless of whether correlations were estimated.
- Fix a bug in S3 method
hypothesis
related to the calculation of Bayes-factors for point hypotheses. - User-defined covariance matrices that are not strictly positive definite for numerical reasons should now be handled correctly.
- Fix problems when a factor is used as fixed effect and as random effects grouping variable at the same time thanks to Ulf Koether.
- Fix minor issues with internal parameter naming.
- Perform additional checking on user defined priors.
- Allow for sampling from all specified proper priors in the model.
- Compute Bayes-factors for point hypotheses in S3 method
hypothesis
.
- Fix a bug that could cause an error for models with multiple grouping factors thanks to Jonathan Williams.
- Fix a bug that could cause an error for weighted poisson and exponential models.
- Implement the Watanabe-Akaike Information Criterion (WAIC).
- Implement the
||
-syntax for random effects allowing for the estimation of random effects standard deviations without the estimation of correlations. - Allow to combine multiple grouping factors within one random effects argument
using the interaction symbol
:
. - Generalize S3 method
hypothesis
to be used with all parameter classes not just fixed effects. In addition, one-sided hypothesis testing is now possible. - Introduce new family
multigaussian
allowing for multivariate normal regression. - Introduce new family
bernoulli
for dichotomous response variables as a more efficient alternative to familiesbinomial
orcategorical
in this special case.
- Slightly change the internal structure of brms to reflect that
rstan
is finally on CRAN. - Thoroughly check validity of the response variable before the data is passed
to
Stan
. - Prohibit variable names containing double underscores
__
to avoid naming conflicts. - Allow function calls with several arguments (e.g.
poly(x,3)
) in the formula argument of functionbrm
. - Always center random effects estimates returned by S3 method
ranef
around zero. - Prevent the use of customized covariance matrices for grouping factors with multiple random effects for now.
- Remove any experimental
JAGS
code from the package.
- Fix a bug in S3 method
hypothesis
leading to an error when numbers with decimal places were used in the formulation of the hypotheses. - Fix a bug in S3 method
ranef
that caused an error for grouping factors with only one random effect. - Fix a bug that could cause the fixed intercept to be wrongly estimated in the presence of multiple random intercepts thanks to Jarrod Hadfield.
- Introduce new methods
parnames
andposterior_samples
for class 'brmsfit' to extract parameter names and posterior samples for given parameters, respectively. - Introduce new method
hypothesis
for classbrmsfit
allowing to test non-linear hypotheses concerning fixed effects. - Introduce new argument
addition
in function brm to get a more flexible approach in specifying additional information on the response variable (e.g., standard errors for meta-analysis). Alternatively, this information can also be passed to theformula
argument directly. - Introduce weighted and censored regressions through argument
addition
of function brm. - Introduce new argument
cov.ranef
in thebrm
function allowing for customized covariance structures of random effects thanks to the idea of Boby Mathew. - Introduce new argument
autocor
in function brm allowing for autocorrelation of the response variable. - Introduce new functions
cor.ar
,cor.ma
, andcor.arma
, to be used with argumentautocor
for modeling autoregressive, moving-average, and autoregressive-moving-average models.
- Amend parametrization of random effects to increase efficiency of the sampling algorithms.
- Improve vectorization of sampling statements.
- Fix a bug that could cause an error when fitting poisson models while
predict = TRUE
. - Fix a bug that caused an error when sampling only one chain while
silent = TRUE
.
- New S3 class
brmsfit
to be returned by thebrm
function. - New methods for class
brmsfit
:summary
,print
,plot
,predict
,fixef
,ranef
,VarCorr
,nobs
,ngrps
, andformula
. - Introduce new argument
silent
in thebrm
function, allowing to suppress most ofStan
's intermediate output. - Introduce new families
negbinomial
(negative binomial) andgeometric
to allow for more flexibility in modeling count data.
- Amend warning and error messages to make them more informative.
- Correct examples in the documentation.
- Extend the README file.
- Fix a bug that caused problems when formulas contained more complicated function calls.
- Fix a bug that caused an error when posterior predictives were sampled for
family
cumulative
. - Fix a bug that prohibited to use of improper flat priors for parameters that have proper priors by default.
- Initial release version