Added functions for computing model averaging weights using Bayesian model averaging (BMA), pseudo-BMA, pseudo-BMA+ (pseudo-BMA with the Bayesian bootstrap), and stacking. Also added a function for generating samples from the ensemble of posterior distributions based on the computed weights.
Added a vignette demonstrating model averaging methods and ensemble inference.
Added functions for sampling from the posterior distributions of several survival models, including accelerated failure time (AFT) models, the piecewise exponential (PWE) model, and the mixture cure rate model with a PWE component for the non-cured popluation (CurePWE). For all survival model implementations, multiple historical data sets are now stacked into a single combined data set for model fitting.
Added functions to compute the log marginal likelihood for the AFT, PWE, and CurePWE models under various prior specifications.
Added the implementation of stratified power prior.
Updated the implementation of generalized linear models (GLMs) to support computation of the pointwise log-likelihood matrix, consistent with the survival model implementations. This enables downstream use in model comparison and averaging procedures, such as computing pseudo-BMA and stacking weights, as well as estimating the expected log predictive density (ELPD).
Added lower and upper bounds for the probability of being exchangeable in the LEAP implementation for GLMs.
Modified the output of the glm.rmap() function to
include the updated mixture weight for the posterior density under the
meta-analytic predictive (MAP) prior.
Added functions for computing the logarithm of the marginal likelihood of a GLM under all priors implemented in the package.
Updated the implementation of commensurate prior to be fully Bayesian.
Updated the implementation of robust meta-analytic predictive
prior (RMAP) from using a Gaussian mixture model to approximate the MAP
prior to computing the updated mixture weights based on marginal
likelihoods. Specifically, we removed glm.rmap.bhm() and
glm.rmap.bhm.approx() functions. The posterior samples from
using the RMAP now can be obtained via calling glm.rmap()
function directly.
Added function for sampling from the posterior distribution of a GLM under a normal/half-normal prior.
Added the vignette “AIDS_Progression”.
Updated Stan files for NAPP and NPP by eliciting priors on logit(a0) instead of a0.
Added two data sets: E1684 and E1690.
Fixed bugs in checking input offset.list for
glm.leap().
Fixed bugs in computing normalizing constants for normal and
Gamma models using glm.npp.lognc().
Fixed bugs in renaming/reordering output variables in
glm.bhm().