Movatterモバイル変換

rifttable 0.7.1

• Housekeeping for CRAN release

rifttable 0.7.0

• Breaking changes:
  • Require base R pipe|> and thus R >= 4.1.
  • Make theid variable identifying clustered observationswithin the same individual a globalrifttable() option forthe entire data set, not only for specific estimators.
• New functionality:
  • Expand input checks to missing values in time/event variables, tomissing effect modifiers for joint models and their levels, tononexistent custom estimators, and to empty input data sets.
  • Addtype = "sum" estimator.
• Internal and bug fixes:
  • Cover entire package by unit tests.
  • Return exposure consistently as acharacter.
  • Let thedesign acceptweight in additiontoweights.
  • Require{risks}>= 0.4.3.
  • Use modern tidyselect, anonymous functions, and code style.

rifttable 0.6.3

• New functionality:
  • Provide easier interface to code competing events by directlyprovidingevent = "event_variable@Event_Type_One" in thedesign. If multiple event types are present, estimatecumulative incidence and differences/ratios of cumulative incidence in acompeting-event setting.
  • Allow for clustered observations in survival data, e.g., multiplerows per person.
  • Estimate ratios of survival and cumulative incidence, i.e.,x-year risk ratios. Use MOVER estimation for confidenceintervals of both ratios and differences in survival and cumulativeincidence by default.
  • Support weights, e.g., inverse-probability weights, directly via aweights argument in thedesign. Weightedestimates are currently available for many survival estimators:type = "cuminc","surv", their differences andratios (e.g.,"cumincdiff"), and"hr". This isa breaking change for Cox models (type = "hr"), whereprovidingweights in thearguments list nowgenerates an error.
• Bug fixes:
  • Allow for@ in factor levels for atable1_design().
  • rt_gt(): Output knitr-formatted tables forGitHub-flavored markdown also in Quarto.qmd, similar to.Rmd.
  • Better handling of edge cases, e.g., ratios of 0, when roundingestimates.
• Expanded documentation
  • Restructure site.
  • Separate documentation of estimators by outcome type.
  • New FAQs on confidence levels, reference levels, custom functions,and joint models.

rifttable 0.6.2

• New functionality:
  • Add overall argumentexposure_levels to let usercontrol handling of missing exposure levels (NA) or factorswith empty levels as the exposure.
  • type = "geomean" for geometric means.
• Documentation: Expand FAQs.
• Internal and bug fixes:
  • Considerexposure ortrend of"" as missing, andstratum = "" as nosubsetting by theeffect_modifier, instead of subsetting toeffect modifier being an empty string. Input check that a stratum mustbe provided for joint models and strata are not empty.
  • Consider missingtype as"blank".
  • Do not add empty rows/columns iftype2 has emptyresults for some cells or if only atrend variable and noexposure is given.
  • Rounding works even if result vector contains strings (e.g., noestimate).
  • More safeguards for all-NAoutcomevariables. More input checks.
  • Do not warn about non-0/1 outcomes inlog-linear models for ratios of continuous variables.
  • Add initial set of unit tests.

rifttable 0.6.1

• New functionality:
  • Cox models (type = "hr") allow forweights, clustering, androbust standarderrors.
  • Argumentratio_digits_decrease: By default, decreasenumber of decimal digits shown for ratios by 1 digit for ratios > 3and by 2 digits for ratios > 10. Leads to rounded ratios andconfidence intervals of1.23,3.4, and11.
  • rt_gt() now indents the first column and appliesmarkdown formatting to it by default.
• New FAQ vignette.
• Bug fixes:
  • Binary outcomes returnedNA instead of0in unstratified tables with all-null outcome.
  • type = "maxfu" ignoreddigits anddiff_digits.
  • Allow for differentexposure (strata labels) andarguments in one table.
  • Show unstratified estimate ifexposure is"", not just forNA.
• Internal:
  • rt_gt(): suppress randomid of gt tablesto keep git diff slim.
  • Keep variables.event,.outcome, etc.available under their original names.
  • Require {risks} >= 0.4.2.
  • Examples load thebreastcancer dataset from the riskspackage.

rifttable 0.6.0

• Breaking changes:
  • to is set to", " by default, instead of"-" for ratio variables and" to " fordifference variables
  • Custom functions are now directly called via thetypevariable, following a restructuring of all estimation functions withgreater flexibility.
  • design$type no longer accepts additional arguments,such as time points. Supply list instead viadesign$arguments.
  • Suppression of strata with sparse or re-identifiable data withdesign$nmin now differentiates between counts of totalobservations or outcomes, depending on estimator.
• New functiontable1_design(): Generate design of adescriptive “Table 1.”
• Newoutcome option"variable@level" forcategorical variables that displayslevel as a binaryoutcome. Used bytable1_design().
• Support unstratified tables displaying the trend/linear slope(trend variable in thedesign) without anexposure.
• More customization:
  • rifttable(reference = ...): Label for the referencecategory.
  • design$ci: Width of confidence intervals.
  • design$na_rm: Omitting observations with missingoutcome data.
  • design$arguments: Flexibly passing along any argumentto estimation functions.
• New vignette describing all estimators.
• Internal:
  • Drop dependency on R >= 4.1 and native pipe.
  • Require{risks}>= 0.4.0.
  • Remove dependency on {labelled} package.
  • The {gt} and {quantreg} packages are now optional as softdependencies.
  • Compatible with {dplyr} 1.1.0, {tidyselect} 1.2.0

rifttable 0.5.0

• khsmisc::table2() “graduated” into its own package. See{khsmisc} Changelogfor earlier versions.
• Addbreastcancer() dataset
• Use R >= 4.1 native pipe,|>
• Remove RMTL estimators