Changelog
Source:NEWS.md
TreeMineR 1.1.1
CRAN release: 2026-08-28
- Corrected the introductory vignette, which was based on results from before the fixes in 1.1.0: it now reports the correct number of significant clusters found in the example data, explains why several cuts can share the same P-value, and explains why some cuts have an infinite risk ratio (RR).
TreeMineR 1.1.0
CRAN release: 2026-08-25
Fixed Monte Carlo P-value calculation in
TreeMineR(). Simulated maximum log-likelihood ratios (LLRs) equal to the observed LLR are now counted as at least as extreme as the observed value (>=instead of>), as required for a valid Monte Carlo P-value. Previously, ties were excluded, which inflated the family-wise type-I error rate, particularly for sparse count data with many possible ties (#9).Fixed
calc_llr()so that nodes in which every individual is exposed (or every individual is unexposed) no longer get an undefined (NaN) log-likelihood ratio. Such nodes were previously dropped from both the simulated reference distribution and the reported results, which could bias P-values (#9).Added a check to
TreeMineR()which now throws an error ifpis not greater than 0 and smaller than 1.p = 0orp = 1made the log-likelihood ratio degenerate (Inf) for all nodes containing exposed, respectively unexposed, individuals.Monte Carlo simulations in
TreeMineR()now preserve the correlation between related cuts. Exposure is simulated once per individual and re-aggregated through the tree, rather than resampling every cut independently, so a parent cut’s simulated count is now mechanically consistent with its children’s, as it is in the observed data (#10).Added a check which now throws an error if
datadoes not contain any rows, and iftreecontains leafs that are duplicated across different branches. Both previously caused an uninformative error from internaldata.tableoperations.
TreeMineR 1.0.4
CRAN release: 2026-02-27
Fixed issue in the example dataset: The example dataset included individuals with non-constant exposure status, i.e., individuals who were at the same time exposed and unexposed. This should not be allowed. The example dataset has been updated accordingly.
Added a check to the TreeMineR function which now throws an error if exposure status is not constant within ids.
TreeMineR 1.0.2
CRAN release: 2024-08-27
- Added option to
TreeMineR()for returning the test distribution of LLRs used for calculating the P-values. - Fixed issues with the ATC tree included in the package. The previous version of the tree did only include ATC codes with a depth of at least 5 levels. The updated version includes also codes with a lower depth, e.g.,
A02AH, which only has a depth of 4 levels.