Re: [Rd] qcauchy accuracy (PR#7902)

From: <maechler_at_stat.math.ethz.ch>
Date: Wed 01 Jun 2005 - 06:57:18 GMT


>>>>> "Morten" == Morten Welinder <mwelinder@gmail.com> >>>>> on Fri, 27 May 2005 20:24:36 +0200 (CEST) writes:

    Morten> Full_Name: Morten Welinder Version: 2.1.0 OS: src     Morten> only Submission from: (NULL) (216.223.241.212)

    Morten> Now that pcauchy has been fixed, it is becoming     Morten> clear that qcauchy suffers from the same problems.

    Morten>
    Morten> qcauchy(pcauchy(1e100,0,1,FALSE,TRUE),0,1,FALSE,TRUE)

    Morten> should yield 1e100 back, but I get 1.633178e+16.     Morten> The code below does much better. Notes:

    Morten> 1. p need not be finite. -Inf is ok in the log_p     Morten> case and R_Q_P01_check already checks things.

yes

    Morten> 2. No need to disallow scale=0 and infinite     Morten> location.

yes

    Morten> 3. The code below uses isnan and finite directly.     Morten> It needs to be adapted to the R way of doing that.

I've done this, and started testing the new code; a version will be put into the next version of R.

Thank you for the suggestions.

>> double
>> qcauchy (double p, double location, double scale, int lower_tail, int log_p)
>> {
>> if (isnan(p) || isnan(location) || isnan(scale))
>> return p + location + scale;

>> R_Q_P01_check(p);
>> if (scale < 0 || !finite(scale)) ML_ERR_return_NAN;

>> if (log_p) {
>> if (p > -1)
>> lower_tail = !lower_tail, p = -expm1 (p);
>> else
>> p = exp (p);
>> }
>> if (lower_tail) scale = -scale;
>> return location + scale / tan(M_PI * p);
>> }



R-devel@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel Received on Wed Jun 01 17:05:08 2005

This archive was generated by hypermail 2.1.8 : Mon 20 Feb 2006 - 03:21:06 GMT