From: <maj_at_stats.waikato.ac.nz>

Date: Fri, 14 Sep 2007 09:54:09 +1200 (NZST)

R-devel_at_r-project.org mailing list

https://stat.ethz.ch/mailman/listinfo/r-devel Received on Fri 14 Sep 2007 - 15:16:54 GMT

Date: Fri, 14 Sep 2007 09:54:09 +1200 (NZST)

I believe that this may be more appropriate here in r-devel than in r-help.

The normal hazard function, or reciprocal Mill's Ratio, may be obtained in R as dnorm(z)/(1 - pnorm(z)) or, better, as dnorm(z)/pnorm(-z) for small values of z. The latter formula breaks dowm numerically for me (running R 2.4.1 under Windows XP 5.1 SP 2) for values of z near 37.4 or greater.

Looking at the pnorm documentation I see that it is based on Cody (1993) and thence, going one step further back, on Cody (1969). Consulting Cody (1969) I see that the algorithm for pnorm(z) [or actually erf(z)] is actually based on rational function approximations for the reciprocal Mill's ratio itself, as I rather expected.

I wonder if anyone has dug out a function for the reciprocal Mill's ratio out of the pnorm() code? Anticipating the obvious response I don't believe that this would be one of the things I might be good at!

Murray Jorgensen

References

Cody, W. D. (1993)

Algorithm 715: SPECFUN – A portable FORTRAN package of special function
routines and test drivers.

ACM Transactions on Mathematical Software 19, 22–32.

Cody, W. D. (1969)

Rational Chebyshev Approximations for the Error Function
Mathematics of Computation, Vol. 23, No. 107. (Jul., 1969), pp. 631-637.

R-devel_at_r-project.org mailing list

https://stat.ethz.ch/mailman/listinfo/r-devel Received on Fri 14 Sep 2007 - 15:16:54 GMT

Archive maintained by Robert King, hosted by
the discipline of
statistics at the
University of Newcastle,
Australia.

Archive generated by hypermail 2.2.0, at Fri 14 Sep 2007 - 20:41:50 GMT.

*
Mailing list information is available at https://stat.ethz.ch/mailman/listinfo/r-devel.
Please read the posting
guide before posting to the list.
*