Bug fixes:
- Handle case where last observation is an event leading to NaN (thanks @trinhdhk)
Bug fixes:
- Fixed case-cohort sampling example thanks @tagteam
Bug fixes:
- Handle missing data correctly with gee.fit/multiple time points
New features:
- New pseudo module for infinitesimal jackknife that allows for delayed entry/left truncation.
New features:
- Allow id argument in cumincglm and rmeanglm. When id is present, geese.fit will be used instead of glm.fit, with clusters indicated by the id. This is to account for clustered data, for example.
- Informative error message if there is missing data in the censoring model
- Add references to JSS publication
- Remove tidyverse lifecycle badge
New features:
Time vector allowed in cumincglm. Use this to model multiple timepoints simultaneously, and allow for time varying covariate effects using tve() (for time varying effect) in the right side of the formula. Inverse probability of censoring weights are returned by the fitting functions in the list element called “ipcw.weights”. * Small edits and improvements to the documentation.
Bug fixes:
- Check and fix name clashes for reserved variables pseudo.vals, .Tci, and .Ci
- Minor efficiency updates in corrected covariance estimation
- Improve documentation
New features and vignette:
- Methods for computing pseudo observations are now in modules.
- Users can define their own modules for computing pseudo observations
- Option to use survival instead of cumulative incidence in standard models
- Update tests/jackknife-agree.R to do tolerance based comparison rather than floating point comparison
- fix typos and add reference to DESCRIPTION