NEWS

bpcp 1.4.2 (2022-03-11)

changed class(x)=="specificClass" statements to is(x,"specificClass") as requested by CRAN.

included new parmtypes ('cdfratio', 'one.minus.ratio', 'one.minus.cdfratio') for bpcp2samp.
fixed typo in output for bpcp2samp (delete "Null" from Nullparm name).

added new function ‘bpcpfit’ that will produce multiple survival curves with the pointwise confidence intervals for right censored data for different treatment/grouping variables.
a plotstyle can be specified in ‘bpcpfit’ so the output can be plotted in either base R or ggplot

fixed references to Fay, Proschan and Brittain (2013) and Fay and Brittain (2016).
fixed error in lines.kmciLR that occurred when marktime=TRUE but there is no censored values.
edited some examples in man pages

fixed warning in bpcp2samp. The answers were correct, but the warning was due to bad coding. Specifically, the abmm function returned a list with its elements greater than 1 dimensional vectors when length(a)>1 and all(b==0). Thanks to Marcel Wolbers for pointing this out.

fixed error when first observation was censored. This error was introduced in 1.3.0.
update references to Fay and Brittain 2016.

major change in the way confidence intervals at observed failures are handled. See vignette on Discrete data and bpcp. Because it only changes the CIs exactly at the failure times, results from simulations done with continuous failure time distributions will not change.
fixed errors in the way Delta>0 was handled in bpcp.
change getmarks functions so they get censoring marks when there are failures at that time also. This fixes an error in the lines.kmciLR when there was censoring but only at times when there was also failures.
change default color for the survival line in lines.kmciLR, it was gray(1) (i.e., white), I meant for it to be gray(0) (i.e., black).

change error message so that it is more clear when there are negative times.
fix NAMESPACE so that it imports even from packages that are included in the R release

added a seed argument to the bpcpControl. Now the default is to set the seed for the Monte Carlo method so that the same data set will give the same answer.
changed the names of the demo files and added a simulation script for the Fay and Brittain (2015) paper.
added a pdf doc to show different calculation methods on two data sets.
added monotonic option to bpcp function to ensure monotonicity of lower and upper limits in time, and made it the default when using Method of Moments.
added control for passing tol argument to uniroot used in bpcp function when midp=TRUE and nmc=0.
set keyword=internal for internal functions to hide internal function help
updated citation file

-added another option for afterMax argument in the StCI function to handle NA for the last confidence interval estimates. -changed help for sclerosis data set. The event is death not relapse (see Figure 5 of Nash, et al 2007).
change qqbeta to use qbeta if using R Version>=3.1.1. -fixed a bug with bpcp2samp when there is no events in one of the groups. -added comments to kmciFunctions.R file
added mid-p option to the bpcp and bpcp2samp functions. This allows mid-p-like confidence intervals. These are more close to having the coverage equal to the nominal coverage on average over all the possible parameter values, but do not guarantee nominal coverage for any specific parameter value.
fixed StCI so that the output data.frame has variables named 'lower' and 'upper' instead of 'lower 95 CL' and 'upper 95 CL'. Since the latter names would change if the conf.level changed, this was causing unnecessary warnings (for example in kmtestALL).

modify kmgw.calc so that it (and bpcp) will not produce NAs from integer overflow when you have a very large number of observations (thanks to Robert Scott for pointing this out).

modify output for bpcp so that it outputs beta parameters associated with method of moments implementation. This is needed for testing at fixed time points in the two sample case using the bpcp2samp function.
create bpcp2samp function for the two sample test for differences at a fixed time point using the confidence distribution approach.
create fixtdiff function for the two sample test for differences at a fixed time point using asyptotic normal approximations -create mdiffmedian.test function for the two sample test for differences in medians (with no censoring)
edit help for kmciLR and kmci objects.
add conf.level for output to kmciLR objects