Skip to content

Latest commit

 

History

History
91 lines (54 loc) · 2.87 KB

NEWS.md

File metadata and controls

91 lines (54 loc) · 2.87 KB
Raw

personalized 0.2.7

  • Bug fix for GAM-based losses
  • Replaces gbm implementation with xgboost using a custom loss function
  • Adds vignette for estimation of flexible ITRs via xgboost

personalized 0.2.6

  • Various small improvements/bug fixes
  • Added changes to reflect the incoming JSS publication related to this package
  • The DOI in the CITATION is for a new JSS publication that will be registered after publication on CRAN.

personalized 0.2.5

  • Adds utilities for construction of augmentation functions and propensity score fitting functions
  • Adds vignette for multi-category treatments and for usage of augmentation/propensity score utilities
  • Various bug fixes

personalized 0.2.4

  • Added warning for use of Harrell's procedure in high dimensions
  • Changed default value for 'train.fraction' to 0.75 from 0.5 in 'validate.subgroup'
  • Minor improvements to plotting for 'subgroup_validated' objects

personalized 0.2.3

  • Fixes trt factor level reordering issue for plots

personalized 0.2.2

  • Fixes model printing error
  • Improves subgroup.summarize() output

personalized 0.2.1

  • Fixes default argument bug for r-oldrel + windows

personalized 0.2.0

  • Simplified plot labeling
  • Added clarifications to documentation
  • Added customized loss function option
  • Added options for count outcomes via Poisson negative log-likelihood as the loss
  • Added treatment effect calculation based on estimated benefit scores
  • Clarified/improved printing
  • Improved numerical stability of weighted.ksvm

personalized 0.1.5

  • Added printing of subgroup_validated results for quantile cutoffs via the which.quant argument

personalized 0.1.4

  • Added plots of means within treatment groups as the benefit scores are varied
  • Added quantile and median cutpoints as options
  • Fixes subgroup effect calculation to account for weights

personalized 0.1.3

  • Added NSW Study dataset

personalized 0.1.2

  • Added requirement for latest version of glmnet. old versions throw error when efficiency augmentation used

personalized 0.1.1

  • Fixed minor bugs regarding multiple treatment options, match.id

  • Added OWL-type losses: logistic and hinge surrogates (multiple treatment available for logistic surrogate)

  • Added outcome flipping OWL-type losses: logistic and hinge surrogates

  • Added augmentation option for non-continuous outcomes via offset

personalized 0.1.0

  • Added estimation functionality for multiple treatments

  • Updated all plot and summary type functions to properly handle results for multiple treatments

  • Aaron added plotting option for validation objects that allows the user to inspect the distribution of variable selections over the bootstrap or training/test resampling iterations

  • Added more options to printing of validation objects

  • Updated check.overlap() to handle multiple treatments

  • Aaron added match.id argument to allow proper analysis of matched case-control datasets