Skip to contents

Changelog

Source: NEWS.md

jsurvival 1.0.4 (2026-08-07)

No analysis changes; the documentation was overhauled. This release keeps the module version in step with the ClinicoPath suite, which moved to 1.0.4 across all of its modules.

Every survival analysis this module ships — datetimeconverter, multisurvival, oddsratio, outcomeorganizer, singlearm, survival, survivalcont and timeinterval — is unchanged since 1.0.2. The diff against that release touches only version strings, the date: field, and two ignore files (.gitignore, .Rbuildignore, for local scratch directories); filtering those out of R/ and jamovi/ leaves no lines of behavioural change. The release-review work done elsewhere in the suite during this window landed in meddecide and OncoPath and is documented in those modules’ NEWS files.

DESCRIPTION (Version 1.0.4, Date 2026-08-06) and jamovi/0000.yaml (version 1.0.4) agree, which is a prerequisite for the release workflow added in 1.0.2 — it refuses to tag when the two disagree.

Documentation

All 53 files under vignettes/ were audited against jamovi/0000.yaml and the generated wrapper signatures. These articles are published to https://www.serdarbalci.com/jsurvival/articles/; vignettes/ is excluded by .Rbuildignore and there is no VignetteBuilder, so none of this affects R CMD check.

  • Option coverage went from 78% to 95%. Of the 304 options across the eight shipped analyses, 67 were not mentioned anywhere in the documentation; 16 remain. Every example added below was executed against the bundled histopathology data before being written down.
  • survival() had 28 undocumented options, several of them substantial features added between February and July 2026. 08-advanced-topics.Rmd now covers them: weighted log-rank tests (weightedLogRank, survivalTestType, with the Fleming-Harrington family reported alongside the standard test); the seven age-correction options (age_adjustment, age_variable, age_interaction, age_stratified_cox, age_group_cutpoints, age_time_scale, age_standardization with age_standardization_method, age_stratified_km); parametric survival models across eight distributions including flexible splines; calibration curves; restricted cubic splines for non-linearity; bootstrap internal validation; adjusted survival curves; and the REMARK reporting checklist.
  • datetimeconverter was shipped with no documentation at all - 30 of its 32 options went unmentioned. 11-data-preparation.Rmd now covers it as the step before interval calculation: the nine input formats, why naming the format beats auto when a dataset mixes dmy and mdy, the eleven component extractors, and the quality report that names rows which failed to parse rather than letting them become silent NAs. The worked example reports 5 of 6 rows parsed (83.33%) with the failure listed by row number.
  • multisurvival’s model-performance options are documented: optimism-corrected C-index by bootstrap (ci_optimism, ci_optimism_boot), covariate contribution by single-term deletion (compare_models), Brier score and time-dependent AUC (show_survmetrics, survmetrics_timepoints), and adjusted probability summaries (ac_summary).
  • A whole article documented a feature that does not exist. 01-multisurvival-time-dependent-comprehensive.Rmd describes time-dependent covariates in multisurvival, using 17 options that are commented out in jamovi/multisurvival.a.yaml under headings reading “EXPERIMENTAL - will be implemented later” - the time-dependent, frailty and spline groups. Sixty options are commented out there in total. The article is kept as the design specification for that future release and now opens by saying the code is not yet runnable; its chunks were already eval = FALSE. multisurvival_documentation.md tabulates 19 such options and now names them.
  • survivalcont was shown with a padjustmethod argument. No such option exists, or ever has. The passage now says so and documents the options that do control multiple cut-points (multiple_cutoffs, num_cutoffs, cutoff_method). The same article claimed restricted cubic splines “would require additional implementation beyond basic jSurvival”; survival() has had rcs_analysis for some time, and the passage now shows it.
  • Twenty-three articles document analyses jsurvival does not ship, and now say so. alluvialSurvival, comparingSurvival, competingsurvival, coxdiagnostics, datecorrection, groupedforest, jvisr, oneSurvival, powersurvival, simonmakuch, stagemigration, subgroupforest, survivalPower, timeroc, lassocox and jiwillsurvive are all on development or test menu routes in the umbrella ClinicoPath module and reach no user today. Separately, 03-treatment-response.Rmd uses crosstable and summarydata (which ship in ClinicoPathDescriptives) and waterfall (OncoPath), and now names the module to install for those steps. Nothing was deleted.
  • All 48 calls to jsurvival’s own analyses across the 53 files were checked against the generated wrapper signatures; the eight failures were the padjustmethod call and the seven time-dependent-covariate calls described above.

jsurvival 1.0.3 (2026-08-04)

An intermediate suite-wide version bump with no jsurvival content. No analysis, option, output or test in this module changed; the commit propagated the 1.0.3 version string and the package date.

jsurvival 1.0.2 (2026-08-03)

All eight analyses were reviewed for this release. Six had a final pre-release pass checking every reported quantity against an independent reference (survival, survRM2, riskRegression, epiR, logistf, lubridate); singlearm and survivalcont were audited earlier in the cycle, and the adjusted-estimation work below spans multisurvival and singlearm together.

The estimators were already correct — Cox hazard ratios, Kaplan-Meier medians, RMST, C-index, Brier scores and adjusted survival curves all reproduce their references exactly. What needed fixing was the boundary between estimated and not estimable, cases where two parts of one report answered the same question differently, and several statements the output made about itself.

Fixed

One estimator behind every panel

  • Changing the adjustment method moved the plot but not the tables. ac_method was read in exactly one place — the call to survminer::ggadjustedcurves() inside the plot. The adjusted survival table, the adjusted median table and the accompanying narrative each built their own prediction instead, from a single mean/mode covariate profile. So average and conditional produced byte-identical tables while the plot changed, and nothing said so; survminer documents those as distinct estimands, not display variants. Plot and table could therefore disagree inside one report. All consumers now read a single shared estimator, .adjustedCurveData(). (multisurvival, singlearm)
  • “Observed at risk” and “observed events” in the adjusted tables were whole-cohort numbers wearing a group label. They came from the model’s common risk set, so they were identical for every level of the adjustment variable while the narrative described them as group-specific. The columns are now titled for what they are, and the estimand is stated above the table.
  • “Adjusted Cox Model Results” was a different model from the main table. In one real run the main multivariable table showed performance_status as a factor with two rows and 13 likelihood-ratio degrees of freedom, while the adjusted table showed it as one continuous row with 12 df — two models fitted, both displayed, nothing saying so. One code path honoured the column’s factor type and the other coerced it to a score. Both now derive from the same fitted object. (multisurvival)
  • Competing-risk detection ignored a recoded outcome column. The guard read multievent && analysistype == "compete", which is blind to the outcomeorganizer hand-off — a recoded Censored/Event/Competing column arrives already coded 0/1/2 with multievent left off, which is the entire point of that column. The 0/1/2 vector then flowed into code that assumes a 0/1 indicator. Detection is now driven by the status vector itself. (survivalcont, and the same fix across the family)

singlearm and survivalcont

  • singlearm results did not invalidate when the outcome recoding, landmark or time units changed. analysistype, the four event-level mappings, uselandmark, landmark, timetypeoutput and timetypedata were missing from the clearWith of the median table, survival table, person-time table, baseline-hazard table and every plot.
  • singlearm multi-event level validation missed unset levels. The check collected the four level options with c(), which silently drops NULLs, so a missing assignment went undetected instead of being reported. Each level is now checked individually.
  • singlearm named the wrong confidence-interval method for the median. The note said conf.type = 'log-log'; survfit()’s default is 'log' — Greenwood’s variance on the log scale. No interval changed, only the label.
  • singlearm described its survival table’s event rate inaccurately. “Event rate by this timepoint” is events within the interval expressed as a proportion of the initial cohort, and now says so.
  • singlearm cutpoints are now parsed and support-checked. .resolveCutpoints() handles malformed input and .supportedCutpoints() refuses time points the data cannot support — including a zero-width person-time boundary, which yields no useful rate. Y-axis limits outside 0–1, or inverted, now produce an actionable message instead of a broken plot.
  • survivalcont stranded its outputs when cut-point search ran. .run() was restructured so results are populated regardless of the findcut path, with deleteRows() guards against double-population, notice reset between runs, an explicit error below ten events, and a validation halt that stops rather than continuing on invalid input.

Estimated vs. not estimable, and output that overstated itself

  • survival reported survival probabilities beyond the observed follow-up. The 1/3/5-year table is built with summary(fit, times =, extend = TRUE), which carries the last Kaplan-Meier estimate forward indefinitely. With the default cutpoints (12, 36, 60) and a cohort followed for two years it printed a “60-month survival” — with a confidence interval — computed from zero patients at risk: 0.0% in one group and 4.5% (0.8–25.7%) in the other. Cutpoints past a group’s last observation are now omitted, with a note naming them and each group’s longest follow-up. This is the same rule .calculateRMST() already applied; it is now applied here too.
  • survival aborted entirely when a cutpoint was mistyped. as.numeric("abc") is NA, and summary.survfit() then stops with times contains missing values — taking the median table, the Cox output and every plot down with it, for one typo in a free-text box. Non-numeric and negative cutpoints are now ignored with a note, duplicates are dropped, and an empty box falls back to the documented default.
  • survival tables did not always refresh when the option driving them changed. cutp was missing from the 1/3/5-year table’s clearWith, and analysistype, the four event-level mappings and the landmark options were missing from four tables. jmvcore’s Table$addRow() appends with no duplicate-key check, so an uncleared table keeps the old rows alongside the new.
  • survival told users to judge significance by eye from two confidence intervals. The interpretation panel stated “Non-overlapping confidence intervals suggest significant differences” as a decision rule, alongside “Use for patient stratification and treatment decisions”. Overlapping intervals do not imply the absence of a difference; the text now points to the log-rank test and drops the treatment-decision claim.
  • multisurvival reported no joint test for a multi-degree-of-freedom interaction. The interaction table gives one row per coefficient, each a 1-df Wald test. For a 3-level × 2-level interaction that is two rows — p = 0.077 and p = 0.726 — and neither answers “is the effect modified by this variable at all?” The joint 2-df test over the same model is p = 0.154. It is now reported alongside the rows (Wald, agreeing with the likelihood-ratio test to three decimals: χ² = 3.7426 / 2 df / p = 0.1539 against 3.7466 / 2 / 0.1536).
  • multisurvival person-time could show figures from a previous covariate set. Person-time is computed on the complete-case set across all selected variables, so adding one continuous covariate with missing values moved the overall row from 134 events / 8235.5 person-time to 94 / 6005.47 — but contexpl was absent from that table’s clearWith and the table never cleared its rows. Both are fixed.
  • oddsratio printed a non-estimable odds ratio as a precise number. Under (quasi-)separation glm’s IRLS halts wherever the iteration limit leaves it, and the table rendered 118848049086800030859264.00 (0.00-Inf, p=1.000) — an odds ratio of 1.19 × 10²³ shown to two decimal places. The cell now reads not estimable, with a strong warning naming the variable and pointing to Firth penalized regression. The rule keys on the confidence interval being unbounded, so a large but genuinely estimable odds ratio is untouched.
  • oddsratio reported diagnostic metrics with no uncertainty. Sensitivity, specificity and the likelihood ratios were bare point estimates: 63.8% from 20 patients and from 2000 read identically. They now carry Clopper-Pearson exact intervals for the proportions and the log method (Simel et al. 1991) for the likelihood ratios — reproducing epiR::epi.tests() bit-identically (maximum absolute difference 0 across 4,156 two-by-two tables). The arithmetic is adapted from epiR (GPL ≥ 2, credited in the source) rather than imported, so jsurvival does not gain an epidemiology package as a dependency for one call.
  • oddsratio named the wrong confidence-interval method. The footnote said “Wilson score”; epi.tests() defaults to method = "exact", which is Clopper-Pearson. Wilson differs in the third decimal.
  • datetimeconverter warned “Implausible Dates Detected” on every successful conversion. A bare format() inside the package resolves to jmvcore::format, which ignores the "%Y" format string; as.integer() of the whole datetime string then yielded the epoch seconds (1710498030 for 2024-03-15) against an upper bound that had become 21033, so every ordinary date compared as out of range. A warning that fires on 100%-successful conversions is worse than none — it trains users to ignore the one signal that a wrong format has mis-parsed their dates. The check now discriminates correctly: silent on 2022–2024, firing on 1850 and 2100.
  • datetimeconverter showed large numbers in scientific notation — and parsed them that way. The same masking made scientific = FALSE a no-op, so a Unix epoch appeared in the preview’s “Original Value” column as 1.7e+09. In the numeric fallback branch that string is also what gets handed to the date parser, turning a valid number into an unparseable date.
  • datetimeconverter documentation contradicted the tool. The glossary gave Excel serial 45000 as “May 18, 2023”; it is 15 March 2023, which is what the converter itself returns. The Excel entry now also explains the 1899-12-30 origin and the 1900 leap-year quirk.
  • datetimeconverter claimed its numeric output was timezone-independent. A date with no time is midnight in the selected zone, so 2024-01-15 becomes 1705276800 under UTC and 1705266000 under Europe/Istanbul. With the default (“system”) two collaborators on different machines get different numbers from the same file. The option text now says so and tells you to select UTC when the value must be reproducible.
  • timeinterval could write the calculated column back with no row mapping. self$data holds only the columns the analysis asked for, so it has one column whenever the same variable is chosen as both start and end date; subsetting without drop = FALSE collapsed it to a vector, rownames() returned NULL, and setRowNums(NULL) then misaligned the written column.
  • timeinterval extreme-value filtering deleted genuine follow-up. The rule is “> multiplier × 99th percentile”, which only orders correctly for a positive 99th percentile. In a cohort where 99% of patients enter and exit on the same day the threshold is 0, so every non-zero interval counted as extreme and the handful of real follow-ups were dropped. The filter is now skipped in that case, with the reason stated under “Filters applied”.
  • outcomeorganizer lost the competing-risk flag when no competing event was observed. The recoded column was exported as a character vector, and jmvcore derives a column’s levels from the values present — so a competing-risks run on a cohort without other-cause deaths shipped a column declaring only Censored/Event. Downstream, survival and multisurvival identify this interchange format by requiring all three declared levels, so the hand-off was not recognised: the analysis stopped with “Event Level is not selected”, or ran as plain Kaplan-Meier if the user picked “Event” to get past it. The column is now exported as a factor declaring Censored/Event/Competing whether or not all three occur, which also puts the levels in the order the 0/1/2 codes imply rather than alphabetically.

Changed

  • Numerical verification tests now verify numbers. The six test-*-verification.R files computed reference fits — and in one case wrote the right answer in a comment — then asserted only that an object came back or that a table had rows. They now compare the rendered values against survival, survRM2, riskRegression, epiR, logistf, stats::t.test, poisson.test and lubridate, and pin every fix above with a regression test.
  • survival: four options that did nothing were removed. hazard_plots, parametric_extrapolation, extrapolation_time and parametric_diagnostics had stub backends and permanently hidden results, but the jamovi UI compiler adds a control for every declared option — so they appeared as live checkboxes, one of them ticked by default. Saved .omv files referencing them still load.

Added

  • Automated GitHub release (.github/workflows/release.yaml). A push to the default branch touching DESCRIPTION or jamovi/0000.yaml cross-checks the two version strings, refuses to proceed if they disagree, and — if the tag does not already exist — tags v<version> and publishes a release whose notes are the matching section of this file.

jsurvival 1.0.0 (2026-07-13)

First stable release

  • Fixed both odds-ratio validation paths so empty or invalid data produce clear analysis-level errors instead of an R6 method failure.
  • Replaced non-structural named HTML entities with Unicode and converted fragmented translated output to complete placeholder-based sentences.
  • Normalized checkbox labels to noun phrases, removed stale sas invalidation keys, and hid unfinished survival-tree controls until the feature is wired end to end.
  • Trimmed 00refs.yaml to the 22 citations used by the eight analyses and replaced broad package imports with selective imports.
  • Synchronized DESCRIPTION, jamovi metadata, and CITATION.cff; removed the orphaned survival-power event handler.

jsurvival 0.0.47 (2026-07-05)

Bug Fixes

  • Fixed a crash in competing-risks analysis on jamovi installs. cmprsk::cuminc() is called by the survival, single-arm, continuous-survival, and multivariable-survival analyses, but cmprsk was missing from the package Imports. Because jamovi installs only a package’s Imports, competing-risks analysis failed on a clean install with “there is no package called ‘cmprsk’”. cmprsk is now declared.
  • Declared digest, pec, and rpart.plot (previously used via :: but undeclared). Moved the optional machine-learning survival backends (randomForestSRC, xgboost, survivalsvm, SurvMetrics) to Suggests so they degrade gracefully when not installed.

jsurvival 0.0.46 (2026-07-04)

This release consolidates versions 0.0.32.62 through 0.0.46 into a single entry. Headline themes: a major expansion of the univariate Survival module toward REMARK-compliant prognostic reporting (age-adjusted analysis, a weighted log-rank test family, calibration assessment, non-linearity testing, and bootstrap internal validation); Cox interaction / effect-modification terms in Multivariable Survival; Firth penalized logistic regression in Odds Ratio; and a unified multi-tier HTML notice system across modules. Minimum jamovi version raised to 2.7.27.

New Statistical Features

Univariate Survival (survival)

  • Age-adjusted survival analysis — new option group controlling age adjustment (age_adjustment, age_variable):
    • age_interaction — test an age x group interaction (new ageInteractionTable with coefficient, HR, SE, z, p-value).
    • age_stratified_cox and ageAdjustedCoxTable comparing unadjusted vs. age-adjusted hazard ratios side by side.
    • age_time_scale — fit a Cox model using age as the time scale (new ageTimeScaleTable).
    • age_standardization with age_standardization_method (indirect/SMR or direct) producing an ageStandardizationTable (observed/expected deaths, SMR with 95% CI).
    • age_stratified_km and age_group_cutpoints for age-stratified Kaplan-Meier curves (ageStratifiedKMPlot).
    • adjusted_curves — covariate-adjusted survival curves (adjustedCurvesPlot).
    • Interpretation panels: ageAdjustedInterpretation, ageAdjustedExplanation, ageTimeScaleInterpretation, ageStandardizationInterpretation.
  • Weighted log-rank test family — new weightedLogRank option with survivalTestType choices: logrank, gehan_breslow, tarone_ware, peto_peto, and fleming_harrington; results in weightedLogRankTable (test, rho, chi-square, df, p-value, weighting) plus a weightedLogRankExplanation panel. Pairwise group comparisons now honor the selected rho weighting.
  • Calibration assessment — calibration_curves with calibration_timepoint and calibration_ngroups; outputs calibrationTable, per-group calibrationGroupTable (predicted vs. observed with CIs), calibrationPlot, and calibrationInterpretation.
  • Non-linearity assessment (restricted cubic splines) — rcs_analysis with rcs_variable and rcs_knots; outputs rcsTestTable (model, df, log-likelihood, AIC, LR chi-square, p-value, conclusion), rcsPlot, and rcsInterpretation.
  • Bootstrap internal validation — bootstrapValidation with bootstrapValN resamples; bootstrapValidationTable reports apparent, optimism, and optimism-corrected metrics with a bootstrapValidationExplanation.
  • REMARK reporting checklist — remark_checklist option renders a remarkChecklist HTML panel for prognostic-marker reporting.
  • Parametric survival scaffolding — UI options added (use_parametric, parametric_distribution covering exponential, Weibull, log-normal, log-logistic, gamma, generalized gamma, Gompertz, and Royston-Parmar spline; spline_knots, spline_scale, parametric_covariates, parametric_extrapolation, extrapolation_time, parametric_diagnostics, compare_distributions, parametric_survival_plots, hazard_plots). The parametric backend remains disabled/experimental in this release.

Multivariable Survival (multisurvival)

  • Cox interaction / effect-modification terms — new interactions option (type Terms) that crosses variables already chosen as explanatory or continuous-explanatory predictors.
    • New interactionTest table (interaction HR with 95% CI and p-value) and subgroupHR table (within-subgroup hazard ratios by moderator level).
    • New pure, unit-testable helper module R/multisurvival-interactions.R (term mapping, formula construction, moderator/subgroup summaries), separated from the R6 backend for maintainability.
    • New jamovi/js/multisurvival.events.js model-builder events that populate the interaction predictor pool from the explanatory/contexpl boxes and prune stale terms.

Odds Ratio (oddsratio)

  • Firth penalized logistic regression — new usePenalized option (via logistf) to reduce small-sample bias and handle separation, with profile-likelihood confidence intervals and automatic fallback to standard logistic regression when logistf is unavailable.
  • New predictorLevel option to set the positive level of the predictor.

Enhanced Existing Modules

Multivariable Survival (multisurvival)

  • Proportional-hazards testing (ph_cox) now defaults to on, surfacing global and per-covariate Schoenfeld residual statistics via survival::cox.zph (aligned with REMARK reporting guidance).

DateTime Converter (datetimeconverter)

  • Added a unified HTML notices panel for validation and conversion messaging.

Notices & Messaging System

  • Introduced a four-tier structured HTML notice system (errors, strongWarnings, warnings, infoMessages) with .addHtmlMessage() / .initializeMessageOutputs() helpers, wired into survival, multisurvival, oddsratio, and outcomeorganizer. Message outputs are reset at the start of each run to prevent accumulation across runs.

Module Removals

  • Removed the Date/DateTime Validator (datevalidator) module (backend, header, and all .a/.r/.u.yaml files deleted).

Package Infrastructure

  • Raised the minimum jamovi application version (minApp) from 1.8.1 to 2.7.27.
  • Added internal helper library R/diagnostichelpers.R with diagnostic-accuracy functions (sensitivity, specificity, PPV/NPV with optional prevalence/Bayes adjustment, positive/negative likelihood ratios, diagnostic odds ratio, Youden’s J).
  • Added internal helper library R/survivalPower_distributions.R implementing Weibull, log-normal, and piecewise-exponential parameterizations and expected-events calculations (Lachin & Foulkes reference).
  • Expanded dataset roxygen documentation in R/data.R (histopathology, melanoma, longitudinal, and stage-migration test datasets).
  • Reworked reproducible-syntax generation (asSource) and hardened R-string escaping across module functions. R/utils.R gains new formula helpers (.asSurvivalFormula(), .escapeVariableNames(), .buildSurvivalFormula()) and %notin%/%!in% operators so that variable names containing special characters produce valid Syntax-mode output.
  • Hardened error handling in Time Interval Calculator (timeinterval) with additional jmvcore::reject() guards for invalid or missing dates and unsupported date formats.
  • Updated bibliographic references in jamovi/00refs.yaml.

jsurvival 0.0.32.60 (2025-12-28)

Major Updates

New Modules

  • Date Validator: Comprehensive date and datetime validation tool.
    • Validates and diagnoses messy date/datetime formats using multiple methods (datefixR, anytime, lubridate).
    • Provides detailed audit tables and quality assessment reports.

Re-introduced and Enhanced Modules

  • Outcome Organizer: Re-introduced with significant enhancements.
    • Improved data cleaning and label handling.
    • Contextual validation for different survival analysis types (OS, Cause-specific, etc.).
    • Diagnostic messages and glossary for educational support.
  • Time Interval: Re-introduced as a comprehensive calculator.
    • Robust time interval calculations with landmark analysis support.
    • Person-time calculation for epidemiological studies.
    • Data quality assessment for time intervals.

Enhancements

Survival Analysis

  • Parametric Models: Improved handling of covariates and formula construction.
  • Plots: Added logic to skip certain plots (e.g., PH assumption, residuals) for competing risk analysis to avoid errors.
  • Diagnostics: Enhanced model summaries and p-value calculations.

General

  • Notice System: Improved notice generation and handling across modules.
  • Documentation: Updated documentation for all modules.

jsurvival 0.0.31.84 (2025-10-03)

New Features

Educational Explanations

  • Added showExplanations option to survival, survivalcont, multisurvival, and oddsratio modules
  • Built-in HTML explanations for key analysis concepts and statistical methods
  • Context-sensitive educational content to help users understand their analyses

Single Arm Survival Enhancements

  • Clinical Analysis Presets: Pre-configured settings for common study types
    • Overall survival analysis (most common)
    • Disease-free survival analysis
    • Treatment effectiveness study
    • Post-surgical outcomes
    • Custom analysis (advanced users)
  • Guided Setup Mode: Step-by-step guidance for users new to survival analysis
  • Enhanced person-time calculations and natural language summaries

Stage Migration Analysis

  • Advanced TNM staging validation and trend analysis
  • Improved robustness and explanations for stagemigration module
  • Multifactorial analysis capabilities
  • Granular table controls and debug outputs

Survival Analysis Module

  • Enhanced educational explanations for univariate survival
  • Improved Cox regression output and interpretation
  • Additional visualization options

Continuous Survival Analysis

  • Educational explanations for cut-point analysis
  • Enhanced optimal threshold determination methods
  • Improved output tables and visualizations

Multivariable Survival

  • Educational explanations for multivariable models
  • Enhanced model diagnostics and output

Odds Ratio Analysis

  • Educational explanations for odds ratio interpretation
  • Improved forest plots and summary tables

Module Removals

  • Removed outcomeorganizer module (functionality integrated into other modules)
  • Removed timeinterval module (functionality integrated into other modules)

Bug Fixes and Improvements

  • Enhanced diagnostic test utility functions
  • Improved tree analysis for survival models
  • Better handling of missing data
  • Updated documentation across all modules

Documentation

  • Updated package documentation and vignettes
  • Enhanced inline help and tooltips
  • Improved example datasets and use cases

jsurvival 0.0.3.90 (2024-07-31)

Major Changes

  • Initial implementation of educational explanations system
  • Refactored analysis modules for better user experience
  • Enhanced natural language summary generation

New Features

  • Stage migration analysis for cancer staging studies
  • Advanced survival model options
  • Diagnostic test utility functions

jsurvival 0.0.3.0 and earlier

For changes in earlier versions, please refer to the git commit history.