Skip to content

NEWS.md

Latest commit

 

History

History
87 lines (45 loc) · 3.52 KB

NEWS.md

File metadata and controls

87 lines (45 loc) · 3.52 KB

version 0.1.4

safer handling of posterior structural zi probability in 0/0 form

version 0.1.3

Defaults pseudo lambda LRT now uses use_cont

version 0.1.2

Lots of changes. Primarily analysis is now centered around a pvlrt object that is created when the function pvlrt() is used. Various summary, plotting, printing methods for pvlrt are created.

version 0.1.1

  • Changes in lrt_zi_poisson():
  • safer null statistics evaluation via tryCatch()
  • better print and summary methods for pvlrt objects
  • functions for extracting p_values, lrstat, zi probabilities from pvlrt object. (Documentation needs to be added.)

version 0.1.0

  • Changes in lrt_zi_poisson():

    • removes the argument skip_null_omega_estimation. Null omega estimation is not performed as it is not necessary.

version 0.0.2.3

  • Changes in lrt_zi_poisson():

    • used analytic expression for pseudo lrt statistic for lambda, and a reduced algebraic expression for llik of omega. Both speed up computations.

    • Replaced omega_est_vec by omega_vec. omega_est_vec is now deprecated: will throw a warning if used instead of omega_vec.

    • Added the argument test_drug_idx to specify which columns to test.

version 0.0.2.2

  • add a secondary optim() call in gamma smoothed omega estimation function

version 0.0.2.1

  • add in the 10 missing rows (with names containing string 'total') in statin

version 0.0.2.0

  • added options "test_omega", "pval_ineq_strict", "skip_null_omega_estimation", and "use_gamma_smooth_omega" to lrt_zi_poisson()

    • test_omega: performs pseudo LRT for the zi parameter omega
    • pval_ineq_strict: determines whether to use >= or > in the definition of the p-values of the test
    • skip_null_omega_estimation: determines whether or not to skip estimation of the omega on the null tables for lambda test. Defaults to TRUE, in which case the test effectively reduces to a parametric bootstrap pseudo-likelihood ratio test. NOTE: this changes the default behavior of the function. In previous versions the same null omegas were used in all null tables, which can currently be achieved by setting skip_null_omega = FALSE
    • use_gamma_smooth_omega: determines whether or not to use a gamma prior (smoothing) while estimating the omegas. Defaults to FALSE. NOTE: this changes the default behavior of the function. In previous versions a gamma prior was always used, which in current version can be achieved by setting use_gamma_smooth_omega = FALSE

  • added 4 new data sets: statin (all pts), statin1491, statin46 and gbca

  • added function r_contin_table_zip() to generate random contingency tables for adverse effect (across rows) and drug (across columns) combinations given row and column marginal totals, embedded signals, and possibly with zero inflation

version 0.1.1.2

  • added option 'grouped_est_omega' (logical argument) in lrt_zi_poisson to specify whehter or not do grouped estimation of omega in extended LRT
  • added a check in lrt_zi_poisson and lrt_vanilla_poisson to ensure that drug_class_idx has the same columns as the input contin_table

version 0.0.1.1

  • updated estimation of omega's in lrt_zi_poisson, to reflect grouping structure among drugs, if present, as provided through drug_class_idx.

version 0.0.1

  • Changed internals of lrt_zi_poisson and lrt_vanilla_poisson:

    • Instead of storing random null tables and LRT statistics through lapply, they are now computed inside a for loop, and p-value is updated on the fly.

  • Removed pbapply from import; added progress

version 0.0.0.9000

The first development version of the package
⚡