Skip to content

NEWS.md

Latest commit

 

History

History
41 lines (37 loc) · 2.27 KB

NEWS.md

File metadata and controls

41 lines (37 loc) · 2.27 KB
Raw

svyTable1 0.14.1.9000

New features

  • svydesign_build() builds a complex-survey design from plain column names with input validation and clear error messages, and supports subpopulation ("domain") analysis the correct way -- by subsetting the design rather than pre-filtering rows, which would break subgroup standard errors.
  • addint() and addintlist() gain ci_method = c("delta", "mover"). The new "mover" option uses the MOVER interval of Zou (2008) doi:10.1093/aje/kwn104 for RERI, which generally has better coverage than the symmetric Wald interval. The default ("delta") is unchanged.
  • addintlist() now returns a Scale column ("OR (logistic)", "HR (Cox)", "RR (log-binomial)", "RR (Poisson)", or "ratio") that makes the effect-measure scale of RERI/AP/S explicit.
  • svykmplot() gains an se argument. se = FALSE draws the survival curves without the (slow) survey-weighted confidence bands.
  • New "Getting Started" vignette giving an end-to-end SPPH 604 workflow, and a package-level help page (?svyTable1) indexing the functions by purpose.

Improvements and bug fixes

  • svygof() now attaches the decile groups to the existing design via update() instead of rebuilding a bare svydesign, so it preserves finite population corrections and calibration and works on replicate-weights designs (it previously errored on them).
  • svyAUC() builds its confidence interval on the logit scale (so it stays within [0, 1]) and uses the design degrees of freedom instead of a fixed critical value; removed redundant internal computation.
  • svycoxph_CE_mi() default time_var/status_var corrected to "stime"/"status".
  • svytable1() now errors on a zero-row design instead of returning a placeholder table, and no longer over-suppresses cells whose design effect is below 1.
  • Documentation: method references with DOIs added to DESCRIPTION; every exported function's \value now names its output class and meaning; the additive-interaction help notes that RERI/AP/S approximate risk-ratio measures only when the outcome is rare.
  • Examples and vignettes are now self-contained on the bundled nhanes_mortality data and no longer require the NHANES package.
  • Added a real test suite (replacing placeholder smoke tests).