1.6.7
-edited man pages for power2x2 and ss2x2 to state that paired=TRUE is not allowed, and direct users to powerPaired2x2
-added stop statement to give an error for paired=TRUE for power2x2 and ss2x2
1.6.6
-fixed power2x2 when approx=T and strict=T (thanks to John Lawrence for pointing out the problem).
-updated references
1.6.5
-fixed uncondExact2x2 (and functions that call it, like boschloo) so othat it does not give a wrong p-value when the choose function goes to Inf. It gave mistakenly gave p=1. Now the function uses the lchoose function to avoid that problem (thanks to Sven Garbade for finding the bug).
1.6.4
-change author field in DESCRIPTION file
- fixed exact2x2 so that midp works when paired=TRUE
-add mcnemarExactDP for exact McNemar-type tests (i.e., exact conditional tests with paired data) with confidence intervals on the difference in proportions
-added powerPaired2x2 for the exact McNemar-type test power
-fixed typos in man page for power2x2 (thanks to Eric Leifer).
-added powerBorr to get power for the boundary-optimized rejection region (BORR) test
1.6.3.1
-No changes to the functions in exact2x2. Just changed test scripts that called exact.test from the Exact R package. Version 2.0 of Exact changed the argument in exact.test from 'interval' to 'np.interval', so that created an error for our test functions. Now test works as long as you use Exact version 2.0 or greater.
1.6.3
-fix bug in borrTest. There was a warning statement that gave an error when borrControl$orderFunc=NULL (which was the default!).
1.6.2
- add borrTest (boundary optimal rejection region test), an unconditional exact test for the two-sample case for comparing test to control where all controls are expected to fail.
1.5.2
- add exact unconditional tests (see help for uncondExact2x2) including Boschloo test (see boschloo function). Includes plotting function for ordering function (plotT), vignettes on details and validation. Sally Hunsberger helped with writing these functions.
- add extra Appendix to Exact McNemar vignette to explain equation 1 better (thanks to Christian Holm Hansen for pointing out the lack of clarity there and for comments on the new explanation, and for Michael Proschan for help with the new Appendix A).
1.4.1
- fix Table for Case-Control studies in exact McNemar vignette (thanks to Eric Leifer for pointing out the error).
- fix ss2x2 so that it does not run forever in certain situations. The problem was that the exact calculation assumes that the starting value underestimates the power. Now we check to make sure that is true, and if not change the starting value. (Thanks to Wolfgang Hartman for pointing out this error).
1.4.0
- fix binomMeld.test when values are at the extremes for the difference test. See code for changes. Also created doc file giving the details.
- update citations
1.3.0
- add functions to calculate exact sample size and power
- add function to calculate the melded binomial confidence interval. Provides confidence intervals for difference, ratio, and odds ratio designed to guarantee coverage.
1.2-0.0
- fix McNemar's test (exact2x2 paired=TRUE) so that if both off diagonal values are 0, then gives p-value of 1 instead of error.
- moved vignettes to vignette directory
1.1-1.0
- fix problem for exact2x2 function when input uses `x,y'-style input and both x and y are factors with 2 levels,
but one of x or y has only one level observed.(Thanks to Hamilton Gross for finding this error).
1.1-0.0
- fix problem with the exact2x2CI function. Now calculates this in a much faster and more accurate way. Previously,
it gave incorrect intervals for the tsmethod="minlike" in some cases with large counts. See comments in
interval function within exact2x2CI function of R code. (Thanks to Nan Bing for finding this error).
1.0-1.1
- update CITATION file
1.0-1
- update reference to Fay (2009) to reflect print version, Fay (2010)
- add plot option to exact2x2 function to plot p-value function
- change exact2x2Plot and internal function so that method is now tsmethod
1.0-0
- add exact McNemar's test (paired=TRUE), for that the package now depends on the exactci package
- add envir=parent.frame to do.call in fisher.exact, without it gave error if name of first input is "y"
- do same as above for blaker.exact
- update reference to Fay (2009), add citation file
0.9-3
- fix problem with exact2x2CI, gave wrong confidence interval for some 2x2 tables
- fix conf.int=FALSE to work correctly for 'blaker' and 'minlike' methods, still ignored for 'central'
- fix p-value for blaker test to be always less than or equal to 1
- fix blaker.exact and fisher.exact to work correctly when y!=NULL
- minor editing of some man pages
- minor editing of some code for clarity
0.9-1 First release