crr.fit caused by extracting column when
input data set is a tibble (#1)QHScrnomo::prostate.dat) instead of using the
data() functionrms 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).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))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.addOffset4ModelFrame, nomo2.crr,
pred2.crr, pred3.crr,
predictDesign)failcode in
pred.ci to failcode=1 to match
cmprsk::crr and QHScrnomo::crr.fittime in
tenf.crr and predict.crr to
time=NULL; also set default arguments for g,
cuts, xlab, ylab arguments in
groupcitrace 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.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.
x, simply pass 1-x into the function
call.a and
b arguments (for the intercept and slope, respectively), as
well as ab, xlim, and ylim
arguments to adjust the graph accordingly.predict.cmprsk, sas.cmprsk,
pred.ci, tenf.crr, groupci,
nomogram.crrtestthatoldUnclass utility
(internal) function. It’s definition was
function(x) unclass(x), so we just use
unclass(x) instead.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.nomogram.crr, changed some arguments:
NULL default for failtime to
make it more clear that this is a required argumentnomogram.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)
R/ for
Getlim, predictDesign (contains
addOffset4ModelFrame), VarcovAdding 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: