fisher.test() gives wrong confidence interval (PR#4019)

Kurt Hornik · 2003-08-30T12:44:51Z

>>>>> jerome writes: > The problem occurs when the sample odds ratio is Inf, such as in the > following example. Given the fact that both upper bounds of the two 95% > confidence intervals are Inf, I would have expected that the two lower > bounds be equal, but they aren't. > x x > # [,1] [,2] > #[1,] 9 0 > #[2,] 4 2 > rbind("two.sided.95CI"=fisher.test(x)$conf.int, > "greater.95CI"=fisher.test(x,alt="greater")$conf.int) > # [,

Kurt Hornik

Sat, Aug 30, 2003 5:44 AM

Right.  When going for a confidence interval that is to include the
parameter estimate and this is already at the boundary of the parameter
range, then there is no point finding mass even further away as there
obviously cannot be any.

I changed the code (r-devel) to

        CINT <- switch(alternative,
                       less = c(0, ncp.U(x, 1 - conf.level)),
                       greater = c(ncp.L(x, 1 - conf.level), Inf),
                       two.sided = {
                           if(ESTIMATE == 0)
                               c(0, ncp.U(x, 1 - conf.level))
                           else if(ESTIMATE == Inf)
                               c(ncp.L(x, 1 - conf.level), Inf)
                           else {
                               alpha <- (1 - conf.level) / 2
                               c(ncp.L(x, alpha), ncp.U(x, alpha))
                           }
                       })

which seems to fix the problem.

Thanks,

-k

fisher.test() gives wrong confidence interval (PR#4019)

Thread (3 messages)