QHScrnomo 3.0.2

• HTML documentation issue

QHScrnomo 3.0.1

• Fixed issue in crr.fit caused by extracting column when input data set is a tibble (#1)

QHScrnomo 3.0.0

Major changes

• Lazily-load internal package data set (i.e., QHScrnomo::prostate.dat) instead of using the data() function
• Changed dependency structure to only use the minimal necessary (see DESCRIPTION)
  • rms is now the only package that QHScrnomo “Depends” on (i.e., it will be loaded when this package is loaded)
  • cmprsk and Hmisc have been moved to the “Imports” field and are accessed internally with the pkg::func format. Therefore, users should not expect these packages to load with QHScrnomo (though they are required to be installed).
• Changed the output of sas.cmprsk to print the base failure probability (1 - exp(-cumsum(f$bfitj)) when the time argument is used instead of the survival probability (exp(-cumsum(f$bfitj))
• Removed the pred.crr function from the package. This function was not previously exported or used, and was highly duplicative of predict.cmprsk. Nevertheless, it is no longer in the package.
• Removed .Rd files for functions that are not exported (addOffset4ModelFrame, nomo2.crr, pred2.crr, pred3.crr, predictDesign)
• Changed the default argument for failcode in pred.ci to failcode=1 to match cmprsk::crr and QHScrnomo::crr.fit
• Set the default argument for time in tenf.crr and predict.crr to time=NULL; also set default arguments for g, cuts, xlab, ylab arguments in groupci
• Added the trace argument to tenf.crr to optionally trace the process in the console. It still does it by default, but previously it always traced it with no option to turn it off.
• In groupci, the ci argument allows the user to control whether we work on the scale of survival or failure (default) probabilities. When the argument was set to FALSE, this would apply a 1-value operation to both the within-group cumulative incidence estimate (to convert to survival) and the group-level mean (of x). However, x is supposed to be an arbitrary continuous variable that we can assess calibration for, not necessarily a probability. Therefore, the change made in this version is to only apply the 1-value transformation to the within-group cumulative incidence and not the group level mean for x.
  • To retain the same behavior when a probability is entered for x, simply pass 1-x into the function call.
  • Users can also use the new a and b arguments (for the intercept and slope, respectively), as well as ab, xlim, and ylim arguments to adjust the graph accordingly.

Minor improvements or bug fixes

• Added more comprehensive error-handling throughout with more informative messaging:
  • Checking for the appropriate input objects in (almost) all functions
  • Checking for the appropriate time point inputs in predict.cmprsk, sas.cmprsk, pred.ci, tenf.crr, groupci, nomogram.crr
  • Other miscellaneous warnings and/or errors (see unit tests)
• Added a unit testing suite via testthat
• Changed, edited, and/or rewrote most (if not all) function documentation text (.Rd files) to be more descriptive and provide more context
• Removed definition and usage of the oldUnclass utility (internal) function. It’s definition was function(x) unclass(x), so we just use unclass(x) instead.
• In groupci, the user is now able to supply xlab = "" as an argument. Also removed the usage of single in the function definition in favor of double as the former is only to be used in the context of .C/.Fortran (per documentation), which is not the case here.
• In nomogram.crr, changed some arguments:
  • Removed the NULL default for failtime to make it more clear that this is a required argument
• Majority of internal/utility functions are used in nomogram.crr and nomogram.mk6. Moved short utility functions to be housed in R/nomogram.R directly (nomo2.crr, is.category, getOldDesign, Design.levels, value.chk, axisf)
  • Still maintaining separate files in R/ for Getlim, predictDesign (contains addOffset4ModelFrame), Varcov

Historical changes

Adding historical change-tracking here that was included in the previous version’s NEWS file (2.2.0) (in reverse order), but in a Markdown-friendly format:

• 2011-05-23 - fix a bug in predict.cmprsk
• 2011-04-19 - change the random sampling scheme in tenf.crr so that the number of patients in each fold is more balanced
• 2011-04-08 - fix a bug in tenf.crr
• 2011-04-13 - add Newlabels.cmprsk and Newlevels.cmprsk
• 2011-03-01 - add cindex, summary.crr and anova.crr
• 2009-06-03 - fixed a bug in sas.crr
• 2008-12-31 - fixed a bug in crr.fit
• 2008-10-16 - edit crr.fit and tenf.crr to deal with subset options in cph
• 2008-08-04 - modify function ‘pred.ci’ to accomodate when failcode == 1
• 2008-04-02 - add ‘tenf.crr’
• 2008-04-02 - add a check to ‘crr.fit’ if bfitj are very large(infinite), stop giving a error message to check if there is a predictor with large integers( i.e, year of surgery)
• 2008-02-21 - add function ‘pred.crr’,‘pred.ci’ and ‘groupci’
• 2008-03-20 - set version to 0.1-3; add flag ‘ci’ to ‘nomogram.crr’
• 2008-03-17 - capitalize package name as QHScrnomo
• 2008-03-14 - add nomogram.crr to correct lp and set verion number to 0.1-2; add function zzz.R
• 2008-03-11 - start to construct the first version 0.1-1