The hardware and bandwidth for this mirror is donated by METANET, the Webhosting and Full Service-Cloud Provider.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]metanet.ch.
posterior_predict() and pp_check(). Regression
wrappers automatically support continuous density checks, discrete
binned-bar checks, scalar test statistics, predictor-based and
fitted-value calibration scatter checks, and conditional,
population-level, or newly simulated random effects. Custom models can
provide replicated outcomes in a generate block.rtmb_mediation(). A random intercept can be included in one
equation or in multiple equations; random intercepts sharing a grouping
variable are modeled jointly with an estimated correlation matrix.gmc/centering)
and centering within cluster (cwc) to
rtmb_mediation(). Centering is applied to predictor uses
while preserving response variables on their original scale; cluster
means remain user-specified model terms..data binding for accessing the
original object passed to rtmb_model(). Wrapper-generated
setup code now reads matrices and data frames directly, or reads related
inputs such as responses, IDs, covariates, and choice sets from one
named data list. Structural options remain visible as fixed assignments
in the generated code.std = TRUE to the regression wrappers. It reports
post-hoc standardized fixed-effect coefficients as b_std,
using every column of the fixed-effect design matrix, including factor
and interaction columns.summary_mcmc() for summarizing numeric MCMC
arrays arranged as iterations by chains by variables. It provides the
same summary columns and print format as
MCMC_Fit$summary(), with parameter and chain
selection.center_grand_mean() and
center_within_cluster() helpers for generated wrapper code
and hand-written models.print_code() expose data-frame
columns instead of a nested data/formula payload, with explicit formula
preprocessing and readable grand-mean and within-cluster centering
helpers.rtmb_table(x, y, data = ...) so unquoted column
names are resolved from the supplied data frame while preserving
reproducible generated code.print_code() output by assigning nfactors and
ndim values directly to the internal constants
K and D.rtmb_corr() normal priors to use the
model-specific mean_sd and sd_rate aliases and
to add an LKJ prior with lkj_eta = 1 by default.fixed, so elements removed from the free-parameter
map no longer contribute to optimization or sampling target
densities..data, unified named data inputs, structural constants in
setup, and reproducible wrapper-generated code.rtmb_lm(), rtmb_glm(),
rtmb_lmer(), and rtmb_glmer() to resolve bare
variable names from the formula environment when data is
omitted. Formulas using $, [[, or
. continue to require explicit data.sigma_rate and tau_rate in
prior_normal() are now 1 / 5, giving
exponential priors with mean 5, and wrapper-specific aliases are applied
consistently.a ~ exponential(1 / 2) when prior_normal() is
used with IRT models.y_range in mixture
and latent-rank models, including response-specific ranges for
multivariate outcomes.rtmb_fa() wrapper now
avoids reading upper-triangular structural-zero entries of
lower_tri loading matrices and constructs constrained AD
matrices with rtmb_array().rtmb_fa() example to a
one-factor model. Advanced factor-analysis workflows remain covered by
documentation and CI regression checks.exp_mod_normal_lpdf() and diffusion_lpdf(),
with sampling syntax support via exp_mod_normal(...) and
obs(RT, Choice) ~ diffusion(...).obs(...) sampling syntax for multivariate
observed values on the left side of ~.setup are more reliably available when building models and
running parallel workers.upgrade_fit() to rebuild saved MCMC, VB, MAP, and
classic fit objects with the currently loaded class definitions,
optionally upgrading their embedded model objects as well.rtmb_vector() and rtmb_array()
tape construction time by automatically reusing an AD seed from model
parameters when available.log_sum_exp(), softmax(), and
log_softmax() work more reliably with RTMB
automatic-differentiation values, including baseline-category patterns
such as softmax(c(0, eta)) inside
rtmb_code().rtmb_vector() and rtmb_array() containers in
loop-filled generated quantities and generated likelihood contributions
where needed.report() handling in transformed and generated
quantities, including namespaced BayesRTMB::report() calls
and wrapper-generated print_code() output.EAP(), MAP(), and
rotation references with the selected best ELBO run while still allowing
explicit chains or best_chains selection.EAP() and MAP() drop their list
wrapper by default when a single parameter is requested, matching the
behavior of estimate().conditional_effects() and simple_effects()
with optimized and classic fits; simple_effects() for
classic fits now also reports df, t value, and
Pr.sd_slice and sd_multiplier controls
for conditional and simple effects, including automatic SD slicing for
moderators with many observed values.rhat_summary() for MCMC fits, returning a numeric
R-hat vector with a compact printed summary.to_long() now supports
multiple value columns, list-based column groups, and preserves input
row order by default while still allowing sorted output with
sort = TRUE.rtmb_vector()
and rtmb_array() for model code that needs mutable
RTMB-compatible containers.rtmb_glmer(cwc = list(ID, "all")), hierarchical
lambda in rtmb_mdu(), stronger prior
validation, and more robust handling of non-finite VB optimization
attempts.\dontrun{} examples with \donttest{} where
appropriate.These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.