- clarified one-sided vs two-sided futility boundary for gsdesign functions

- added check for n; since it is for stopping early at least two values should be given
- addressed the warning in the example for calogrank; status had both death and transplant

- fixed the bug in ph2simon/twostage.admissible when minimax is also optimal
- ph2simon and toxbdry help have been made more clear

- changed all dfloat in fortran to dble
- print method for ph2simon now lists admissible designs

- plot and lines method for roc.curve includes the option for precision-recall curve
- roc.curve returns PPV and NPV which are needed for precision-recall curve

- changed class check to use inherits instead of comparing strings

- Added out.ties option to coxphCPE for discrete risk score case

- Added added admissible two stage design function
- Added added futilbdry function for sequential futility stopping
- added 1:r sample allocation in gsdesign functions
- Fixed the warnings messages from deltaAUC.f

- Added a
`NEWS.md`

file to track changes to the package. - deposited development version on GitHub
- created {pkgdown} website and github actions for building site

- Added new function deltaAUC to test for difference in the area under the ROC curves from nested binary regression models.
- power.ladesign does exact null if number of combinations is small
- Added to help of fedesign, ph2simon and ph2single functions

- added init.c to register native routines

- permlogrank bug fix - strata option gave an error. Names of elements in survfit.object changed. (strata instead of ntimes.strata and n a vector).

- added aucVardiTest to compare growth curves

- check for misssing values in jonckheere.test

- bug fix for pselect; nlen > 3 should have been nlen > 2

- use requireNamespace to remove NOTES on functions using survival pkg

- fix name swap in the result of mdrr in v1.0.7

- new function mdrr added to calculate minimum detectable difference in response rates for given average response rate and class proportion

- On rare occasions jonckheere.test gave a p-value bigger than 1.
Sometimes 2
*min(iPVAL, dPVAL, 1) can be larger than 1. Replace with 2*min(iPVAL, dPVAL, 0.5) (Thanks to Drs. Shterev and Owzar of Duke).

- Fixed the bound checks

- Added the option to calculate continuity corrected sample size in the function gsdesign.binomial.

- Fixed two-sided p-value > 1 bug when statistic is exactly its mean.
- Fixed Rd file to address LaTeX warnings.

- integer overflow in djonck for the Jonckheere-Terpstra statistic. replace with pdf calculation using Mark van de Wiel convolution.
- Create seperate help for functions permlogrank and jonckheere.test

- Fix integer overflow because of n0
*n1 == 0 in roc.curve and nn*nd == 0 in roc.area.test - In coxphQuantile eliminate times for which survival probability is 0 or 1 from quantile computation.

# clinfun number changed to 1.0.0 in preparation for R 2.15.

- toxbdry now does the entire Pocock to O’Brien-Fleming range of boundaries. Added references for the method.
- Fixed linebreak in the help for coxphERR.

- Added coxphERR to calculate Heller’s explained relative risk.
- Fixed NaN bug in toxbdry when priority=“alt” is used.

- fixed bug in print.gsdesign for binomial case (p1,p2 instead of pC,pE)
- ROC functions now check that there are at least one each of status=0,1

- fixed fortran code to address gfortran-4.6 warnings
- added ktau a faster implementation of cor(x, y, method=“k”). not in NAMESPACE. Should be called using clinfun:::ktau

- bug fix: roc.area.test integer overflow for large nn*nd.
- use sort function to speed up roc curve and area estimation.

- bug fix: roc.area.test gave NaN as the statistic and p-value when the markers are identical. Changed it to 0.

- bug fix: gsdesign funtions not returning the sample size / # events.

- twostage.inference for umvue, p-value and CI for 2 stage design.

- ph2simon was testing whether dim is null for feasible solution. Replaced with nrow == 0 since it is now possible to have 0 rows.

- Added functions to compute and plot the empirical ROC curve.

- Added functions for the area and permutation tests to compare ROC.
- Checks that min.diff is greater than 0 in pselect.

- Added a non-binding futility boundary to gsdesign

- variable names for returned data.frame in coxphQuantile
- examples in coxphQuantile & coxphCPE use status==2

- check R# clinfun so that coxphCPE works for any # clinfun (see 0.8.9)

- Change coxphCPE to reflect the fact that model.matrix.coxph doesn’t have an intercept term.

- Added the function or2pcase

- Fixed the 0/0 bug in the revised pselect

- Changed Venkat’s affiliation to MSKCC.
- Fixed pselect to calculate the selection probability correctly when only one treatment exceeds the min.resp threshold.

- Changed the Jonckheere-Terpstra statistic such that large value is indicative of increasing group locations and small for decreasing. Function warns that p-value is based on approximation for tied data.

- Added functionality to pselect. It can do unequal sample size for the case of two treatments. min.diff can be specified as a rate instead of number of responses. Output element names changed to be more descriptive.

- Fixed the bug CPS.ssize. call inside used fixed alpha, power & r.
- Fixed the bug in the approximate one-sided p-value of Jonckheere test. wrong tail was used.

- toxbdry allows the error threshold to prioritize when the sample size is too small to have both satisfied.

- Fixed gsdesign to allow for fixed sample designs. Help file fixed.

- New # clinfun with one new function power.lehmann.design