NEWS.md

In jenfb/bkmr: Bayesian Kernel Machine Regression

bkmr 0.2.2

Bug fixes

• Corrected code that produced warning length > 1 in coercion to logical

• Update functions that use deprecated functions from dplyr package

Minor changes

• No longer export the following functions:

• CalcGroupPIPs, CalcWithinGroupPIPs, and CalcPIPs as these should typically be calculated using the function ExtractPIPs

• ComputePostmeanHnew.approx and ComputePostmeanHnew.exact as these should typically be calculated using the function ComputePostmeanHnew

• set_verbose_opts as this is only called internally

• Expanded function documentation by adding example code

bkmr 0.2.1

Bug fixes

• allowable values for starting parameter for r[m] parameters updated as follows

• no longer truncated to a single value (when varsel = FALSE and rmethod = "varying")

• can be equal to 0 (when varsel = TRUE)

• Error no longer generated if starting values for h.hat are not positive

• When checking class of an object, use inherits() instead of class()

bkmr 0.2.0

Major changes

• Added ability to have binomial outcome family by implementing probit regression within kmbayes()

• Removed computation of the subject-specific effects h[i] within kmbayes(), as this is not always desired, and greatly slows down model fitting

• This could still be done by setting the option est.h = TRUE in the kmbayes function

• posterior samples of h[i] can now be obtained via the post-processing SamplePred function; alternatively, posterior summaries (mean, variance) can be obtained via the post-processing ComputePostmeanHnew function

• Added ability to use exact estimates of the posterior mean and variance by specifying the argument method = 'exact' within the post-processing functions (e.g., OverallRiskSummaries(), PredictorResponseUnivar())

Bug fixes

• Fixed PredictorResponseBivarLevels() when argument both_pairs = TRUE (#4)

jenfb/bkmr documentation built on March 26, 2022, 8:30 p.m.