NEWS.md
Changes
Supporting Bayesian hierarchical meta analysis with bias correction using negative control outcomes by adding computeHierarchicalMetaAnalysis()
.
Providing additional example dataset – profile likelihoods from LegendT2dm.
Supporting simulation for Self-Controlled Case Series (SCCS) data. (see createSccsSimulationSettings()
).
Adding support for synthesizing pooled SCCS data.
Introducing Hermite interpolation using a sparse grid with gradients. These approximations will be faster to construct, especially for models with many (nuisance) parameters such as the SCCS.
Changes
Supporting Bayesian adaptive bias correction in sequential analysis by adding fitBiasDistribution()
, sequentialFitBiasDistribution()
and biasCorrectionInference()
functions
Added relevant plotting functions, plotBiasDistribution()
and plotBiasCorrectionInference()
.
Added a vignette on Bayesian adaptive bias correction.
Changes
Supporting adaptive grid in plotLikelihoodFit()
and computeFixedEffectMetaAnalysis()
functions.
Adding visualization of likelihood curves to plotMetaAnalysisForest()
.
Changes
computeBayesianMetaAnalysis()
and approximateLikelihood()
functions.Bugfixes
1 Fixed error when approximating likelihood using grid (parameter to approximate could only be “x”).
Changes
Higher tolerance on skew-normal unit tests to prevent them from failing.
Detecting and removing grid approximations with illegal values before computing meta-analysis.
Changes
Documenting dependency on Java in the SystemRequirements
field of the package DESCRIPTION.
Adding seed
argument to computeBayesianMetaAnalysis()
. Defaults to a constant value for reproducability.
Bugfixes
plotMetaAnalysisForest()
when using grid approximations, including when providing approximations as tibble.Changes
computeBayesianMetaAnalysis()
now outputs ESS.
Checking whether required Java version (8 or newer) is installed.
Changes
Preparing for CRAN release: Adding missing rmarkdown dependency to Suggests.
Added computeConfidenceInterval()
function.
plotMetaAnalysisForest()
function now works with normal and non-normal approximations.
Grid points now evenly spaced on log scale, not HR scale.