https://github.com/cran/brms
Tip revision: a89e1e8a2b20992a0c36f4511f81038f07203e5a authored by Paul-Christian Bürkner on 23 January 2018, 20:48:07 UTC
version 2.1.0
version 2.1.0
Tip revision: a89e1e8
NEWS.Rd
\name{NEWS}
\title{brms News}
\encoding{UTF-8}
\section{CHANGES IN VERSION 2.1.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Fit models on multiple imputed
datasets via \code{brm_multiple} thanks
to Ruben Arslan. (#27)
\item Combine multiple \code{brmsfit} objects
via function \code{combine_models}.
\item Compute model averaged posterior
predictions with method \code{pp_average}. (#319)
\item Add new argument \code{ordinal} to
\code{marginal_effects} to generate special
plots for ordinal models thanks to the idea
of the GitHub user silberzwiebel. (#190)
\item Use informative inverse-gamma priors
for length-scale parameters of Gaussian
processes. (#275)
\item Compute hypotheses for all levels of a
grouping factor at once using argument \code{scope}
in method \code{hypothesis}. (#327)
\item Vectorize user-defined \pkg{Stan} functions
exported via \code{export_functions} using
argument \code{vectorize}.
\item Allow predicting new data in models
with ARMA autocorrelation structures.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Correctly recover noise-free coefficients
through \code{me} terms thanks to Ruben Arslan.
As a side effect, it is no longer possible to define
priors on noise-free \code{Xme} variables directly, but
only on their hyper-parameters \code{meanme} and \code{sdme}.
\item Fix problems in renaming parameters of the
\code{cor_bsts} structure thanks to Joshua Edward Morten. (#312)
\item Fix some unexpected errors when predicting
from ordinal models thanks to David Hervas and
Florian Bader. (#306, #307, #331)
\item Fix problems when estimating and predicting
multivariate ordinal models thanks to David West. (#314)
\item Fix various minor problems in autocorrelation
structures thanks to David West. (#320)
}
}
}
\section{CHANGES IN VERSION 2.0.1}{
\subsection{NEW FEATURES}{
\itemize{
\item Export the helper functions \code{posterior_summary}
and \code{posterior_table} both being used to
summarize posterior samples and predictions.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix incorrect computation of intercepts
in \code{acat} and \code{cratio} models
thanks to Peter Phalen. (#302)
\item Fix \code{pointwise} computation of \code{LOO}
and \code{WAIC} in multivariate models with estimated
residual correlation structure.
\item Fix problems in various S3 methods sometimes
requiring unused variables to be specified in
\code{newdata}.
\item Fix naming of Stan models thanks to Hao Ran Lai.
}
}
}
\section{CHANGES IN VERSION 2.0.0}{
This is the second major release of \code{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 \pkg{brms} 1.0 or higher should remain
fully compatible with \pkg{brms} 2.0.
\subsection{NEW FEATURES}{
\itemize{
\item 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
\code{gaussian} and \code{student} models. All features
supported in univariate models are now also available in
multivariate models. (#3)
\item Specify different formulas for different
categories in \code{categorical} models.
\item Add weakly informative default priors for the
parameter class \code{Intercept} to improve convergence
of more complex distributional models.
\item Optionally display the MC standard error in the
\code{summary} output. (#280)
\item Add argument \code{re.form} as an alias of
\code{re_formula} to the methods \code{posterior_predict},
\code{posterior_linpred}, and \code{predictive_error}
for consistency with other packages making use of
these methods. (#283)
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Refactor many parts of the package to make
it more consistent and easier to extend.
\item Show the link functions of all
distributional parameters in the \code{summary}
output. (#277)
\item Reduce working memory requirements when
extracting posterior samples for use in \code{predict}
and related methods thanks to Fanyi Zhang. (#224)
\item Remove deprecated aliases of functions
and arguments from the package. (#278)
\item No longer support certain prior specifications,
which were previously labeled as deprecated.
\item Remove the depreacted addition term
\code{disp} from the package.
\item Remove old versions of methods \code{fixef},
\code{ranef}, \code{coef}, and \code{VarCorr}.
\item No longer support models fitted with \pkg{brms} < 1.0,
which used the multivariate \code{'trait'} syntax
orginally deprecated in \pkg{brms} 1.0.
\item Make posterior sample extraction in the
\code{summary} method cleaner and less error prone.
\item No longer fix the seed for random number generation
in \code{brm} to avoid unexpected behavior in simulation studies.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Store \code{stan_funs} in \code{brmsfit} objects
to allow using \code{update} on models with user-defined
Stan functions thanks to Tom Wallis. (#288)
\item Fix problems in various post-processing methods
when applied to models with the reserved variable
\code{intercept} in group-level terms thanks to
the GitHub user ASKurz. (#279)
\item Fix an unexpected error in \code{predict} and related
methods when setting \code{sample_new_levels = "gaussian"}
in models with only one group-level effect.
Thanks to Timothy Mastny. (#286)
}
}
}
\section{CHANGES IN VERSION 1.10.2}{
\subsection{NEW FEATURES}{
\itemize{
\item Allow setting priors on noise-free
variables specified via function \code{me}.
\item Add arguments \code{Ksub}, \code{exact_loo}
and \code{group} to method \code{kfold} for
defining omitted subsets according to a
grouping variable or factor.
\item Allow addition argument \code{se}
in \code{skew_normal} models.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Ensure correct behavior of horseshoe
and lasso priors in multivariate models
thanks to Donald Williams.
\item Allow using \code{identity} links on
all parameters of the \code{wiener} family
thanks to Henrik Singmann. (#276)
\item Use reasonable dimnames in the output
of \code{fitted} when returning linear predictors
of ordinal models thanks to the GitHub user atrolle. (#274)
\item Fix problems in \code{marginal_smooths}
occuring for multi-membership models thanks to
Hans Tierens.
}
}
}
\section{CHANGES IN VERSION 1.10.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Rebuild monotonic effects from scratch
to allow specifying interactions with other
variables. (#239)
\item Introduce methods \code{posterior_linpred}
and \code{posterior_interval} for consistency
with other model fitting packages based on
\pkg{Stan}.
\item Introduce function \code{theme_black}
providing a black \pkg{ggplot2} theme.
\item Specify special group-level effects within
the same terms as ordinary group-level effects.
\item Add argument \code{prob} to
\code{summary}, which allows to control the
width of the computed uncertainty intervals. (#259)
\item Add argument \code{newdata} to the
\code{kfold} method.
\item Add several arguments to the \code{plot}
method of \code{marginal_effects} to improve
control over the appearences of the plots.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Use the same noise-free variables
for all model parts in measurement error models. (#257)
\item Make names of local-level terms used
in the \code{cor_bsts} structure more informative.
\item Store the \code{autocor} argument
within \code{brmsformula} objects.
\item Store posterior and prior samples in separate
slots in the output of method \code{hypothesis}.
\item No longer change the default theme of
\pkg{ggplot2} when attaching \pkg{brms}. (#256)
\item Make sure signs of estimates are not dropped
when rounding to zero in \code{summary.brmsfit}. (#263)
\item Refactor parts of \code{extract_draws}
and \code{linear_predictor} to be more consistent
with the rest of the package.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Do not silence the \pkg{Stan} parser
when calling \code{brm} to get informative
error messages about invalid priors.
\item Fix problems with spaces in priors
passed to \code{set_prior}.
\item Handle non \code{data.frame} objects
correctly in \code{hypothesis.default}.
\item Fix a problem relating to the colour
of points displayed in \code{marginal_effects}.
}
}
}
\section{CHANGES IN VERSION 1.9.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Perform model comparisons based on
marginal likelihoods using the methods
\code{bridge_sampler}, \code{bayes_factor},
and \code{post_prob} all powered by the
\pkg{bridgesampling} package.
\item Compute a Bayesian version of R-squared
with the \code{bayes_R2} method.
\item Specify non-linear models for all
distributional parameters.
\item Combine multiple model formulas using
the \code{+} operator and the helper functions
\code{lf}, \code{nlf}, and \code{set_nl}.
\item Combine multiple priors using the
\code{+} operator.
\item Split the \code{nlpar} argument of
\code{set_prior} into the three arguments
\code{resp}, \code{dpar}, and \code{nlpar}
to allow for more flexible prior specifications.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Refactor parts of the package to prepare
for the implementation of more flexible
multivariate models in future updates.
\item Keep all constants in the log-posterior
in order for \code{bridge_sampler} to be
working correctly.
\item Reduce the amount of renaming done
within the \code{stanfit} object.
\item Rename argument \code{auxpar} of
\code{fitted.brmsfit} to \code{dpar}.
\item Use the \code{launch_shinystan} generic
provided by the \pkg{shinystan} package.
\item Set \code{bayesplot::theme_default()}
as the default \pkg{ggplot2} theme when
attaching \pkg{brms}.
\item Include citations of the \pkg{brms}
overview paper as published in the
Journal of Statistical Software.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix problems when calling \code{fitted}
with \code{hurdle_lognormal} models thanks to
Meghna Krishnadas.
\item Fix problems when predicting \code{sigma}
in \code{asym_laplace} models thanks to
Anna Josefine Sorensen.
}
}
}
\section{CHANGES IN VERSION 1.8.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Fit conditional autoregressive (CAR) models
via function \code{cor_car} thanks to the case
study of Max Joseph.
\item Fit spatial autoregressive (SAR) models
via function \code{cor_sar}. Currently works
for families \code{gaussian} and \code{student}.
\item Implement skew normal models via family
\code{skew_normal}. Thanks to Stephen Martin
for suggestions on the parameterization.
\item Add method \code{reloo} to perform exact
cross-validation for problematic observations
and \code{kfold} to perform k-fold cross-validation
thanks to the Stan Team.
\item Regularize non-zero coefficients in the
\code{horseshoe} prior thanks to Juho Piironen
and Aki Vehtari.
\item Add argument \code{new_objects} to various
post-processing methods to allow for passing of
data objects, which cannot be passed via
\code{newdata}.
\item Improve parallel execution flexibility
via the \code{future} package.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Improve efficiency and stability of ARMA models.
\item Throw an error when the intercept is removed
in an ordinal model instead of silently adding
it back again.
\item Deprecate argument \code{threshold} in \code{brm}
and instead recommend passing \code{threshold} directly
to the ordinal family functions.
\item Throw an error instead of a message when
invalid priors are passed.
\item Change the default value of the \code{autocor}
slot in \code{brmsfit} objects to an empty
\code{cor_brms} object.
\item Shorten \pkg{Stan} code by combining
declarations and definitions where possible.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix problems in \code{pp_check}
when the variable specified in argument
\code{x} has attributes thanks to
Paul Galpern.
\item Fix problems when computing fitted
values for truncated discrete models based
on new data thanks to Nathan Doogan.
\item Fix unexpected errors when passing
models, which did not properly initiliaze,
to various post-processing methods.
\item Do not accidently drop the second
dimension of matrices in \code{summary.brmsfit}
for models with only a single observation.
}
}
}
\section{CHANGES IN VERSION 1.7.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Fit latent Gaussian processes of one
or more covariates via function \code{gp}
specified in the model formula (#221).
\item Rework methods \code{fixef}, \code{ranef},
\code{coef}, and \code{VarCorr} to be more flexible
and consistent with other post-processing methods (#200).
\item Generalize method \code{hypothesis} to be
applicable on all objects coercible to a
\code{data.frame} (#198).
\item Visualize predictions via spaghetti
plots using argument \code{spaghetti} in
\code{marginal_effects} and \code{marginal_smooths}.
\item Introduce method \code{add_ic} to
store and reuse information criteria in
fitted model objects (#220).
\item Allow for negative weights in
multi-membership grouping structures.
\item Introduce an \code{as.array} method
for \code{brmsfit} objects.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Show output of \R code in HTML vignettes thanks
to Ben Goodrich (#158).
\item Resolve citations in PDF vignettes thanks
to Thomas Kluth (#223).
\item Improve sampling efficiency for
\code{exgaussian} models thanks to
Alex Forrence (#222).
\item Also transform data points when using argument
\code{transform} in \code{marginal_effects}
thanks to Markus Gesmann.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix an unexpected error in \code{marginal_effects}
occuring for some models with autocorrelation terms
thanks to Markus Gesmann.
\item Fix multiple problems occuring for models with
the \code{cor_bsts} structure thanks to Andrew Ellis.
}
}
}
\section{CHANGES IN VERSION 1.6.1}{
\subsection{NEW FEATURES}{
\itemize{
\item Implement zero-one-inflated beta models
via family \code{zero_one_inflated_beta}.
\item Allow for more link functions in
zero-inflated and hurdle models.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Ensure full compatibility with
\pkg{bayesplot} version 1.2.0.
\item Deprecate addition argument \code{disp}.
}
}
\subsection{BUG FIXES}{
\itemize{
\item 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.
\item Allow ordered factors to be used
as grouping variables thanks to the GitHub
user itissid.
}
}
}
\section{CHANGES IN VERSION 1.6.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Fit finite mixture models using family
function \code{mixture}.
\item Introduce method \code{pp_mixture} to compute
posterior probabilities of mixture component
memberships thanks to a discussion with Stephen Martin.
\item Implement different ways to sample new levels
of grouping factors in \code{predict} and related
methods through argument \code{sample_new_levels}.
Thanks to Tom Wallis and Jonah Gabry for a detailed
discussion about this feature.
\item Add methods \code{loo_predict}, \code{loo_linpred},
and \code{loo_predictive_interval} for computing
LOO predictions thanks to Aki Vehtari and Jonah Gabry.
\item Allow using \code{offset} in formulas
of non-linear and auxiliary parameters.
\item Allow sparse matrix multiplication in
non-linear and distributional models.
\item Allow using the \code{identity} link for
all auxiliary parameters.
\item Introduce argument \code{negative_rt} in
\code{predict} and \code{posterior_predict} to
distinquish responses on the upper and lower
boundary in \code{wiener} diffusion models
thanks to Guido Biele.
\item Introduce method \code{control_params} to
conveniently extract control parameters of the
NUTS sampler.
\item Introduce argument \code{int_conditions} in
\code{marginal_effects} for enhanced plotting of
two-way interactions thanks to a discussion with
Thomas Kluth.
\item Improve flexibility of the \code{conditions}
argument of \code{marginal_effects}.
\item Extend method \code{stanplot} to correctly
handle some new \code{mcmc_} plots of the
\pkg{bayesplot} package.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Improve the \code{update} method to
only recompile models when the \pkg{Stan} code
changes.
\item Warn about divergent transitions when calling
\code{summary} or \code{print} on \code{brmsfit} objects.
\item Warn about unused variables in argument
\code{conditions} when calling \code{marginal_effects}.
\item Export and document several distribution functions
that were previously kept internal.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix problems with the inclusion of offsets
occuring for more complicated formulas thanks to
Christian Stock.
\item Fix a bug that led to invalid Stan code when
sampling from priors in intercept only models thanks
to Tom Wallis.
\item Correctly check for category specific
group-level effects in non-ordinal models thanks to
Wayne Folta.
\item Fix problems in \code{pp_check} when specifying
argument \code{newdata} together with arguments
\code{x} or \code{group}.
\item Rename the last column in the output of
\code{hypothesis} to \code{"star"} in order to avoid
problems with zero length column names thanks to
the GitHub user puterleat.
\item Add a missing new line statement at the end
of the \code{summary} output thanks to Thomas Kluth.
}
}
}
\section{CHANGES IN VERSION 1.5.1}{
\subsection{NEW FEATURES}{
\itemize{
\item Allow \code{horseshoe} and \code{lasso}
priors to be applied on population-level effects
of non-linear and auxiliary parameters.
\item Force recompiling \pkg{Stan} models
in \code{update.brmsfit} via argument
\code{recompile}.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Avoid indexing of matrices in non-linear
models to slightly improve sampling speed.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a severe problem (introduced in version 1.5.0),
when predicting \code{Beta} models thanks to Vivian Lam.
\item Fix problems when summarizing some models
fitted with older version of \pkg{brms} thanks
to Vivian Lam.
\item Fix checks of argument \code{group} in
method \code{pp_check} thanks to Thomas K.
\item Get arguments \code{subset} and \code{nsamples}
working correctly in \code{marginal_smooths}.
}
}
}
\section{CHANGES IN VERSION 1.5.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Implement the generalized extreme value
distribution via family \code{gen_extreme_value}.
\item Improve flexibility of the \code{horseshoe}
prior thanks to Juho Piironen.
\item Introduce auxiliary parameter \code{mu}
as an alternative to specifying effects within
the \code{formula} argument in function
\code{brmsformula}.
\item Return fitted values of auxiliary parameters
via argument \code{auxpar} of method \code{fitted}.
\item Add vignette \code{"brms_multilevel"}, in which
the advanced formula syntax of \pkg{brms} is explained
in detail using several examples.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item 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.
\item Save the version number of \pkg{rstan} in
element \code{version} of \code{brmsfit} objects.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a rare error when predicting \code{von_mises}
models thanks to John Kirwan.
}
}
}
\section{CHANGES IN VERSION 1.4.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Fit quantile regression models via family
\code{asym_laplace} (asymmetric Laplace distribution).
\item Specify non-linear models in a (hopefully) more
intuitive way using \code{brmsformula}.
\item Fix auxiliary parameters to certain values
through \code{brmsformula}.
\item Allow \code{family} to be specified in
\code{brmsformula}.
\item Introduce family \code{frechet} for modelling
strictly positive responses.
\item Allow truncation and censoring at the same time.
\item Introduce function \code{prior_} allowing
to specify priors using one-sided formulas or \code{quote}.
\item Pass priors to \pkg{Stan} directly without
performing any checks by setting \code{check = FALSE}
in \code{set_prior}.
\item Introduce method \code{nsamples} to extract
the number of posterior samples.
\item Export the main formula parsing function
\code{parse_bf}.
\item Add more options to customize two-dimensional surface
plots created by \code{marginal_effects} or \code{marginal_smooths}.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Change structure of \code{brmsformula}
objects to be more reliable and easier to extend.
\item Make sure that parameter \code{nu} never
falls below \code{1} to reduce convergence problems
when using family \code{student}.
\item Deprecate argument \code{nonlinear}.
\item Deprecate family \code{geometric}.
\item Rename \code{cov_fixed} to \code{cor_fixed}.
\item Make handling of addition terms more transparent
by exporting and documenting related functions.
\item Refactor helper functions of the \code{fitted}
method to be easier to extend in the future.
\item Remove many units tests of internal functions
and add tests of user-facing functions instead.
\item Import some generics from \pkg{nlme} instead
of \pkg{lme4} to remove dependency on the latter one.
\item Do not apply \code{structure} to \code{NULL}
anymore to get rid of warnings in \R-devel.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix problems when fitting smoothing terms
with factors as \code{by} variables thanks to
Milani Chaloupka.
\item Fix a bug that could cause some monotonic
effects to be ignored in the \pkg{Stan} code thanks
to the GitHub user bschneider.
\item Make sure that the data of models with
only a single observation are compatible with
the generated \pkg{Stan} code.
\item Handle argument \code{algorithm}
correctly in \code{update.brmsfit}.
\item Fix a bug sometimes causing an error in
\code{marginal_effects} when using family
\code{wiener} thanks to Andrew Ellis.
\item Fix problems in \code{fitted} when applied
to \code{zero_inflated_beta} models thanks to
Milani Chaloupka.
\item Fix minor problems related to the prediction
of autocorrelated models.
\item Fix a few minor bugs related to the backwards
compatibility of multivariate and related models
fitted with \pkg{brms} < 1.0.0.
}
}
}
\section{CHANGES IN VERSION 1.3.1}{
\subsection{NEW FEATURES}{
\itemize{
\item Introduce the auxiliary parameter \code{disc}
('discrimination') to be used in ordinal models.
By default it is not estimated but fixed to one.
\item Create \code{marginal_effects} plots of
two-way interactions of variables that were
not explicitely modeled as interacting.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Move \pkg{rstan} to 'Imports' and \pkg{Rcpp}
to 'Depends' in order to avoid loading \pkg{rstan}
into the global environment automatically.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug leading to unexpected errors
in some S3 methods when applied to ordinal models.
}
}
}
\section{CHANGES IN VERSION 1.3.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Fit error-in-variables models
using function \code{me} in the model formulae.
\item Fit multi-membership models using function
\code{mm} in grouping terms.
\item Add families \code{exgaussian}
(exponentially modified Gaussian distribution)
and \code{wiener} (Wiener diffusion model distribution)
specifically suited to handle for response times.
\item Add the \code{lasso} prior as an alternative
to the \code{horseshoe} prior for sparse models.
\item Add the methods \code{log_posterior},
\code{nuts_params}, \code{rhat}, and \code{neff_ratio}
for \code{brmsfit} objects to conveniently access
quantities used to diagnose sampling behavior.
\item Combine chains in method \code{as.mcmc} using
argument \code{combine_chains}.
\item Estimate the auxiliary parameter
\code{sigma} in models with known standard errors of
the response by setting argument \code{sigma} to
\code{TRUE} in addition function \code{se}.
\item Allow visualizing two-dimensional smooths
with the \code{marginal_smooths} method.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Require argument \code{data} to be explicitely
specified in all user facing functions.
\item Refactor the \code{stanplot} method
to use \pkg{bayesplot} on the backend.
\item Use the \pkg{bayesplot} theme as the default
in all plotting functions.
\item Add the abbreviations \code{mo} and \code{cs}
to specify monotonic and category specific effects
respectively.
\item Rename generated variables in the data.frames
returned by \code{marginal_effects} to avoid potential
naming conflicts.
\item Deprecate argument \code{cluster} and use
the native \code{cores} argument of \pkg{rstan} instead.
\item Remove argument \code{cluster_type} as it is
no longer required to apply forking.
\item Remove the deprecated \code{partial} argument.
}
}
}
\section{CHANGES IN VERSION 1.2.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Add the new family \code{hurdle_lognormal}
specifically suited for zero-inflated continuous responses.
\item Introduce the \code{pp_check} method to perform
various posterior predictive checks
using the \pkg{bayesplot} package.
\item Introduce the \code{marginal_smooths} method to
better visualize smooth terms.
\item Allow varying the scale of global shrinkage
parameter of the \code{horseshoe} prior.
\item Add functions \code{prior} and \code{prior_string}
as aliases of \code{set_prior}, the former
allowing to pass arguments without quotes \code{""}
using non-standard evaluation.
\item Introduce four new vignettes explaining how to fit
non-linear models, distributional models, phylogenetic models,
and monotonic effects respectively.
\item Extend the \code{coef} method to better
handle category specific group-level effects.
\item Introduce the \code{prior_summary} method
for \code{brmsfit} objects to obtain a summary
of prior distributions applied.
\item Sample from the prior of the original population-level
intercept when \code{sample_prior = TRUE} even in models
with an internal temporary intercept used to improve
sampling efficiency.
\item Introduce methods \code{posterior_predict},
\code{predictive_error} and \code{log_lik} as
(partial) aliases of \code{predict}, \code{residuals},
and \code{logLik} respectively.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Improve computation of Bayes factors
in the \code{hypothesis} method to be less
influenced by MCMC error.
\item Improve documentation of default priors.
\item Refactor internal structure of some
formula and prior evaluating functions.
This should not have any user visible effects.
\item Use the \pkg{bayesplot} package as the
new backend of \code{plot.brmsfit}.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Better mimic \pkg{mgcv} when parsing smooth terms
to make sure all arguments are correctly handled.
\item Avoid an error occuring during the prediction
of new data when grouping factors with only a single
factor level were supplied thanks to Tom Wallis.
\item Fix \code{marginal_effects} to consistently
produce plots for all covariates in non-linear models
thanks to David Auty.
\item Improve the \code{update} method to better recognize
situations where recompliation of the \pkg{Stan} code
is necessary thanks to Raphael P.H.
\item Allow to correctly \code{update} the \code{sample_prior}
argument to value \code{"only"}.
\item Fix an unexpected error occuring in many S3 methods
when the thinning rate is not a divisor of the total
number of posterior samples thanks to Paul Zerr.
}
}
}
\section{CHANGES IN VERSION 1.1.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Estimate monotonic group-level effects.
\item Estimate category specific group-level effects.
\item Allow \code{t2} smooth terms based on
multiple covariates.
\item Estimate interval censored data via the
addition argument \code{cens} in the model formula.
\item Allow to compute \code{residuals} also based
on predicted values instead of fitted values.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Use the prefix \code{bcs} in parameter names
of category specific effects and the prefix \code{bm}
in parameter names of monotonic effects (instead
of the prefix \code{b}) to simplify their identifaction.
\item Ensure full compatibility with
\pkg{ggplot2} version 2.2.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug that could result in incorrect
threshold estimates for \code{cumulative} and
\code{sratio} models thanks to Peter Congdon.
\item Fix a bug that sometimes kept distributional
\code{gamma} models from being compiled
thanks to Tim Beechey.
\item Fix a bug causing an error in \code{predict}
and related methods when two-level factors or
logical variables were used as covariates in
non-linear models thanks to Martin Schmettow.
\item Fix a bug causing an error when passing
lists to additional arguments of smoothing
functions thanks to Wayne Folta.
\item Fix a bug causing an error in the
\code{prior_samples} method for models with
multiple group-level terms that refer to the same
grouping factor thanks to Marco Tullio Liuzza.
\item Fix a bug sometimes causing an error when
calling \code{marginal_effects} for weighted models.
}
}
}
\section{CHANGES IN VERSION 1.0.1}{
\subsection{MINOR CHANGES}{
\itemize{
\item Center design matrices inside the Stan code
instead of inside \code{make_standata}.
\item Get rid of several warning messages
occuring on CRAN.
}
}
}
\section{CHANGES IN VERSION 1.0.0}{
This is one of the largest updates of \code{brms} since its
initial release. In addition to many new features,
the multivariate \code{'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 \code{help(brmsformula)} for details of the formula
syntax applied in \pkg{brms}.
\subsection{NEW FEATURES}{
\itemize{
\item 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 \pkg{lme4} syntax.
\item 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.
\item Zero-inflated and hurdle models do not use
multivariate syntax anymore but instead have
special auxiliary parameters named \code{zi} and
\code{hu} defining zero-inflation / hurdle probabilities.
\item Implement the \code{von_mises} family to model
circular responses.
\item Introduce the \code{brmsfamily} function for
convenient specification of \code{family} objects.
\item Allow predictions of \code{t2} smoothing
terms for new data.
\item Feature vectors as arguments for the addition
argument \code{trunc} in order to model varying
truncation points.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Remove the \code{cauchy} family
after several months of deprecation.
\item Make sure that group-level parameter names
are unambiguous by adding double underscores
thanks to the idea of the GitHub user schmettow.
\item The \code{predict} method now returns predicted
probabilities instead of absolute frequencies of
samples for ordinal and categorical models.
\item 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.
\item Colour points in \code{marginal_effects} plots if
sensible.
\item Set the default of the \code{robust} argument
to \code{TRUE} in \code{marginal_effects.brmsfit}.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug that could occur when predicting
factorial response variables for new data.
Only affects categorical and ordinal models.
\item 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.
\item Fix problems when trying to pointwise
evaluate non-linear formulae in
\code{logLik.brmsfit} thanks to Tom Wallis.
\item Ensure full compatibility of the \code{ranef}
and \code{coef} methods with non-linear models.
\item Fix problems that occasionally occured when
handling \pkg{dplyr} datasets thanks to the
GitHub user Atan1988.
}
}
}
\section{CHANGES IN VERSION 0.10.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Add support for generalized additive mixed models
(GAMMs). Smoothing terms can be specified using
the \code{s} and \code{t2} functions in the model formula.
\item Introduce \code{as.data.frame} and \code{as.matrix}
methods for \code{brmsfit} objects.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item The \code{gaussian("log")} family no longer implies
a log-normal distribution, but a normal distribution with
log-link to match the behavior of \code{glm}.
The log-normal distribution can now be specified via
family \code{lognormal}.
\item Update syntax of \pkg{Stan} models to match the
recommended syntax of \pkg{Stan} 2.10.
}
}
\subsection{BUG FIXES}{
\itemize{
\item The \code{ngrps} method should now always
return the correct result for non-linear models.
\item Fix problems in \code{marginal_effects} for
models using the reserved variable \code{intercept}
thanks to Frederik Aust.
\item Fix a bug in the \code{print} method of
\code{brmshypothesis} objects that could lead to
duplicated and thus invalid row names.
\item Residual standard deviation parameters of
multivariate models are again correctly displayed
in the output of the \code{summary} method.
\item Fix problems when using variational Bayes
algorithms with \pkg{brms} while having
\pkg{rstan} >= 2.10.0 installed thanks to the
GitHub user cwerner87.
}
}
}
\section{CHANGES IN VERSION 0.9.1}{
\subsection{NEW FEATURES}{
\itemize{
\item Allow the '/' symbol in group-level terms in
the \code{formula} argument to indicate nested
grouping structures.
\item Allow to compute \code{WAIC} and \code{LOO}
based on the pointwise log-likelihood using argument
\code{pointwise} to substantially reduce memory requirements.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Add horizontal lines to the errorbars in
\code{marginal_effects} plots for factors.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug that could lead to a cryptic error
message when changing some parts of the model
\code{formula} using the \code{update} method.
\item Fix a bug that could lead to an error when
calling \code{marginal_effects} for predictors
that were generated with the \code{base::scale}
function thanks to Tom Wallis.
\item Allow interactions of numeric and categorical
predictors in \code{marginal_effects} to be passed
to the \code{effects} argument in any order.
\item Fix a bug that could lead to incorrect results
of \code{predict} and related methods when called with
\code{newdata} in models using the \code{poly} function
thanks to Brock Ferguson.
\item Make sure that user-specified factor contrasts
are always applied in multivariate models.
}
}
}
\section{CHANGES IN VERSION 0.9.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Add support for \code{monotonic} effects
allowing to use ordinal predictors without
assuming their categories to be equidistant.
\item Apply multivariate formula syntax in categorical
models to considerably increase modeling flexibility.
\item Add the addition argument \code{disp} to define
multiplicative factors on dispersion parameters.
For linear models, \code{disp} applies to the residual
standard deviation \code{sigma} so that it can be
used to weight observations.
\item Treat the fixed effects design matrix as sparse
by using the \code{sparse} argument of \code{brm}.
This can considerably reduce working memory
requirements if the predictors contain many zeros.
\item Add the \code{cor_fixed} correlation structure to
allow for fixed user-defined covariance matrices of the
response variable.
\item Allow to pass self-defined \pkg{Stan} functions
via argument \code{stan_funs} of \code{brm}.
\item Add the \code{expose_functions} method allowing to
expose self-defined \pkg{Stan} functions in \code{R}.
\item Extend the functionality of the \code{update}
method to allow all model parts to be updated.
\item Center the fixed effects design matrix also
in multivariate models. This may lead to increased
sampling speed in models with many predictors.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Refactor \pkg{Stan} code and data generating
functions to be more consistent and easier to extent.
\item Improve checks of user-define prior specifications.
\item Warn about models that have not converged.
\item Make sure that regression curves computed by
the \code{marginal_effects} method are always smooth.
\item Allow to define category specific effects in
ordinal models directly within the \code{formula}
argument.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix problems in the generated \pkg{Stan} code
when using very long non-linear model formulas
thanks to Emmanuel Charpentier.
\item Fix a bug that prohibited to change priors
on single standard deviation parameters
in non-linear models thanks to Emmanuel Charpentier.
\item Fix a bug that prohibited to use nested
grouping factors in non-linear models thanks to
Tom Wallis.
\item Fix a bug in the linear predictor computation
within \code{R}, occuring for ordinal models
with multiple category specific effects. This
could lead to incorrect outputs of \code{predict},
\code{fitted}, and \code{logLik} for these models.
\item Make sure that the global \code{"contrasts"} option
is not used when post-processing a model.
}
}
}
\section{CHANGES IN VERSION 0.8.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Implement generalized non-linear models, which
can be specified with the help of the \code{nonlinear}
argument in \code{brm}.
\item Compute and plot marginal effects using the
\code{marginal_effects} method thanks to the help
of Ruben Arslan.
\item Implement zero-inflated beta models through
family \code{zero_inflated_beta} thanks to the
idea of Ali Roshan Ghias.
\item Allow to restrict domain of fixed effects and
autocorrelation parameters using new arguments
\code{lb} and \code{ub} in function \code{set_prior}
thanks to the idea of Joel Gombin.
\item Add an \code{as.mcmc} method for compatibility
with the \pkg{coda} package.
\item Allow to call the \code{WAIC}, \code{LOO},
and \code{logLik} methods with new data.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Make sure that \pkg{brms} is fully compatible
with \pkg{loo} version 0.1.5.
\item Optionally define the intercept
as an ordinary fixed effect to avoid the
reparametrization via centering of the
fixed effects design matrix.
\item Do not compute the WAIC in \code{summary}
by default anymore to reduce computation time
of the method for larger models.
\item The \code{cauchy} family is now deprecated
and will be removed soon as it often has convergence
issues and not much practical application anyway.
\item Change the default settings of the number of
chains and warmup samples to the defaults of \pkg{rstan}
(i.e., \code{chains = 4} and \code{warmup = iter / 2}).
\item Do not remove bad behaving chains anymore as
they may point to general convergence problems that
are dangerous to ignore.
\item Improve flexibility of the \code{theme} argument
in all plotting functions.
\item Only show the legend once per page, when computing
trace and density plots with the \code{plot} method.
\item Move code of self-defined \pkg{Stan} functions
to \code{inst/chunks} and incorporate them into the
models using \code{rstan::stanc_builder}.
Also, add unit tests for these functions.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix problems when predicting with \code{newdata}
for zero-inflated and hurdle models thanks to Ruben Arslan.
\item Fix problems when predicting with \code{newdata}
if it is a subset of the data stored in a
\code{brmsfit} object thanks to Ruben Arslan.
\item Fix data preparation for multivariate models
if some responses are \code{NA} thanks to Raphael Royaute.
\item Fix a bug in the \code{predict} method occurring
for some multivariate models so that it now always
returns the predictions of all response variables,
not just the first one.
\item Fix a bug in the log-likelihood computation of
\code{hurdle_poisson} and \code{hurdle_negbinomial} models.
This may lead to minor changes in the values obtained by
\code{WAIC} and \code{LOO} for these models.
\item Fix some backwards compatibility issues of models fitted
with version <= 0.5.0 thanks to Ulf Koether.
}
}
}
\section{CHANGES IN VERSION 0.7.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Use variational inference algorithms
as alternative to the NUTS sampler by specifying
argument \code{algorithm} in the \code{brm} function.
\item Implement beta regression models through family
\code{Beta}.
\item Implement zero-inflated binomial models through family
\code{zero_inflated_binomial}.
\item Implement multiplicative effects for family
\code{bernoulli} to fit (among others) 2PL IRT models.
\item Generalize the \code{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.
\item Combine fixed and random effects estimates using
the new \code{coef} method.
\item Call the \code{residuals} method with \code{newdata} thanks
to the idea of Friederike Holz-Ebeling.
\item Allow new levels of random effects grouping
factors in the \code{predict}, \code{fitted}, and \code{residuals}
methods using argument \code{allow_new_levels}.
\item Selectively exclude random effects in the \code{predict},
\code{fitted}, and \code{residuals} methods using
argument \code{re_formula}.
\item Add a \code{plot} method for objects returned by
method \code{hypothesis} to visualize prior and posterior
distributions of the hypotheses being tested.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Improve evaluation of the response
part of the \code{formula} argument to
reliably allow terms with more than one variable
(e.g., \code{y/x ~ 1}).
\item Improve sampling efficiency of models containing
many fixed effects through centering the fixed effects
design matrix thanks to Wayne Folta.
\item Improve sampling efficiency of models containing
uncorrelated random effects specified by means
of \code{(random || group)} terms in \code{formula}
thanks to Ali Roshan Ghias.
\item Utilize user-defined functions in the \pkg{Stan} code
of ordinal models to improve readability as well as
sampling efficiency.
\item Make sure that model comparisons using \code{LOO}
or \code{WAIC} are only performed when models are
based on the same responses.
\item Use some generic functions of the \pkg{lme4}
package to avoid unnecessary function masking. This
leads to a change in the argument order of
method \code{VarCorr}.
\item Change the \code{ggplot} theme in the
\code{plot} method through argument \code{theme}.
\item Remove the \code{n.} prefix in arguments
\code{n.iter}, \code{n.warmup}, \code{n.thin}, \code{n.chains},
and \code{n.cluster} of the \code{brm} function.
The old argument names remain usable as deprecated aliases.
\item Amend names of random effects parameters to simplify
matching with their respective grouping factor levels.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug in the \code{hypothesis} method
that could cause valid model parameters to be falsely
reported as invalid.
\item Fix a bug in the \code{prior_samples} method
that could cause prior samples of parameters
of the same class to be artifically correlated.
\item Fix \pkg{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.
\item Fix a bug in the evaluation of \code{formula} that
could cause complicated random effects terms to be
falsely treated as fixed effects.
\item Fix several bugs when calling the \code{fitted}
and \code{predict} methods with \code{newdata} thanks
to Ali Roshan Ghias.
}
}
}
\section{CHANGES IN VERSION 0.6.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Add support for zero-inflated and hurdle models thanks
to the idea of Scott Baldwin.
\item Implement inverse gaussian models through family
\code{inverse.gaussian}.
\item Allow to specify truncation boundaries of the response
variable thanks to the idea of Maciej Beresewicz.
\item Add support for autoregressive (AR) effects of residuals, which can
be modeled using the \code{cor_ar} and \code{cor_arma} functions.
\item Stationary autoregressive-moving-average (ARMA) effects
of order one can now also be fitted using special covariance matrices.
\item Implement multivariate student-t models.
\item Binomial and ordinal families now support
the \code{cauchit} link function.
\item Allow family functions to be used in the
\code{family} argument.
\item Easy access to various \code{rstan} plotting functions
using the \code{stanplot} method.
\item Implement horseshoe priors to model sparsity in
fixed effects coefficients thanks to the idea of Josh Chang.
\item Automatically scale default standard deviation priors so that
they remain only weakly informative independent on the response scale.
\item Report model weights computed by the \pkg{loo} package
when comparing multiple fitted models.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Separate the fixed effects Intercept from other fixed effects
in the \pkg{Stan} code to slightly improve sampling efficiency.
\item Move autoregressive (AR) effects of the response from
the \code{cor_ar} to the \code{cor_arr} function as the result
of implementing AR effects of residuals.
\item Improve checks on argument \code{newdata} used in the
\code{fitted} and \code{predict} method.
\item Method \code{standata} is now the only way to extract data
that was passed to \pkg{Stan} from a \code{brmsfit} object.
\item Slightly improve \pkg{Stan} code for models
containing no random effects.
\item Change the default prior of the degrees of freedom of the
\code{student} family to \code{gamma(2,0.1)}.
\item Improve readability of the output of method \code{VarCorr}.
\item Export the \code{make_stancode} function to give users
direct access to \pkg{Stan} code generated by \pkg{brms}.
\item Rename the \code{brmdata} function to \code{make_standata}.
The former remains usable as a deprecated alias.
\item Improve documenation to better explain differences in
autoregressive effects across R packages.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug that could cause an unexpected error
when the \code{predict} method was called with \code{newdata}.
\item Avoid side effects of the \code{rstan} compilation
routines that could occasionally cause R to crash.
\item Make \pkg{brms} work correctly with \pkg{loo} version 0.1.3
thanks to Mauricio Garnier Villarreal and Jonah Gabry.
\item Fix a bug that could cause WAIC and LOO estimates to be slightly
incorrect for \code{gaussian} models with \code{log} link.
}
}
}
\section{CHANGES IN VERSION 0.5.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Compute the Watanabe-Akaike information criterion (WAIC) and
leave-one-out cross-validation (LOO) using the \pkg{loo} package.
\item Provide an interface to \pkg{shinystan} with
S3 method \code{launch_shiny}.
\item New functions \code{get_prior} and \code{set_prior} to make
prior specifications easier.
\item Log-likelihood values and posterior predictive samples can now
be calculated within R after the model has been fitted.
\item Make predictions based on new data using S3 method \code{predict}.
\item Allow for customized covariance structures of
grouping factors with multiple random effects.
\item New S3 methods \code{fitted} and \code{residuals}
to compute fitted values and residuals, respectively.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Arguments \code{WAIC} and \code{predict} are removed
from the \code{brm} function, as they are no longer necessary.
\item New argument \code{cluster_type} in function \code{brm}
allowing to choose the cluster type created by the parallel package.
\item Remove chains that fail to initialize while sampling
in parallel leaving the other chains untouched.
\item Redesign trace and density plots to be faster and more stable.
\item S3 method \code{VarCorr} now always returns covariance
matrices regardless of whether correlations were estimated.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug in S3 method \code{hypothesis} related
to the calculation of Bayes-factors for point hypotheses.
\item User-defined covariance matrices that are not
strictly positive definite for numerical reasons should
now be handled correctly.
\item Fix problems when a factor is used as fixed effect
and as random effects grouping variable at the same time
thanks to Ulf Koether.
\item Fix minor issues with internal parameter naming.
\item Perform additional checking on user defined priors.
}
}
}
\section{CHANGES IN VERSION 0.4.1}{
\subsection{NEW FEATURES}{
\itemize{
\item Allow for sampling from all specified proper
priors in the model.
\item Compute Bayes-factors for point hypotheses
in S3 method \code{hypothesis}.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug that could cause an error for models with
multiple grouping factors thanks to Jonathan Williams.
\item Fix a bug that could cause an error for
weighted poisson and exponential models.
}
}
}
\section{CHANGES IN VERSION 0.4.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Implement the Watanabe-Akaike Information Criterion (WAIC).
\item Implement the \code{||}-syntax for random effects allowing
for the estimation of random effects standard deviations
without the estimation of correlations.
\item Allow to combine multiple grouping factors within one random
effects argument using the interaction symbol \code{:}.
\item Generalize S3 method \code{hypothesis} to be used with
all parameter classes not just fixed effects.
In addition, one-sided hypothesis testing is now possible.
\item Introduce new family \code{multigaussian} allowing
for multivariate normal regression.
\item Introduce new family \code{bernoulli} for dichotomous
response variables as a more efficient alternative to families
\code{binomial} or \code{categorical} in this special case.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Slightly change the internal structure of brms to reflect that
\pkg{rstan} is finally on CRAN.
\item Thoroughly check validity of the response variable
before the data is passed to \pkg{Stan}.
\item Prohibit variable names containing double underscores
\code{__} to avoid naming conflicts.
\item Allow function calls with several arguments
(e.g. \code{poly(x,3)}) in the formula argument of
function \code{brm}.
\item Always center random effects estimates returned
by S3 method \code{ranef} around zero.
\item Prevent the use of customized covariance matrices for
grouping factors with multiple random effects for now.
\item Remove any experimental \pkg{JAGS} code from the package.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug in S3 method \code{hypothesis}
leading to an error when numbers with decimal places
were used in the formulation of the hypotheses.
\item Fix a bug in S3 method \code{ranef} that caused
an error for grouping factors with only one random effect.
\item Fix a bug that could cause the fixed intercept to be
wrongly estimated in the presence of multiple random intercepts
thanks to Jarrod Hadfield.
}
}
}
\section{CHANGES IN VERSION 0.3.0}{
\subsection{NEW FEATURES}{
\itemize{
\item Introduce new methods \code{parnames} and
\code{posterior_samples} for class 'brmsfit' to extract
parameter names and posterior samples for given parameters,
respectively.
\item Introduce new method \code{hypothesis} for class \code{brmsfit}
allowing to test non-linear hypotheses concerning fixed effects.
\item Introduce new argument \code{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 the
\code{formula} argument directly.
\item Introduce weighted and censored regressions through argument
\code{addition} of function brm.
\item Introduce new argument \code{cov.ranef} in the \code{brm}
function allowing for customized covariance structures of
random effects thanks to the idea of Boby Mathew.
\item Introduce new argument \code{autocor} in function brm
allowing for autocorrelation of the response variable.
\item Introduce new functions \code{cor.ar}, \code{cor.ma},
and \code{cor.arma}, to be used with argument \code{autocor}
for modeling autoregressive, moving-average,
and autoregressive-moving-average models.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Amend parametrization of random effects to
increase efficiency of the sampling algorithms.
\item Improve vectorization of sampling statements.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug that could cause an error when fitting
poisson models while \code{predict = TRUE}.
\item Fix a bug that caused an error when sampling only one chain
while \code{silent = TRUE}.
}
}
}
\section{CHANGES IN VERSION 0.2.0}{
\subsection{NEW FEATURES}{
\itemize{
\item New S3 class \code{brmsfit} to be returned by
the \code{brm} function.
\item New methods for class \code{brmsfit}:
\code{summary}, \code{print}, \code{plot}, \code{predict},
\code{fixef}, \code{ranef}, \code{VarCorr},
\code{nobs}, \code{ngrps}, and \code{formula}.
\item Introduce new argument \code{silent} in the
\code{brm} function, allowing to suppress most of
\pkg{Stan}'s intermediate output.
\item Introduce new families \code{negbinomial} (negative binomial)
and \code{geometric} to allow for more flexibility
in modeling count data.
}
}
\subsection{OTHER CHANGES}{
\itemize{
\item Amend warning and error messages to make
them more informative.
\item Correct examples in the documentation.
\item Extend the README file.
}
}
\subsection{BUG FIXES}{
\itemize{
\item Fix a bug that caused problems when formulas contained
more complicated function calls.
\item Fix a bug that caused an error when posterior predictives were
sampled for family \code{cumulative}
\item Fix a bug that prohibited to use of improper flat priors
for parameters that have proper priors by default.
}
}
}
\section{CHANGES IN VERSION 0.1.0}{
\itemize{
\item Initial release version
}
}