R devel - Apr 2004 - pbinom accuracy (PR#6757)

Full_Name: Morten Welinder
Version: snapshot
OS: 
Submission from: (NULL) (65.213.85.129)


There is really no good reason for throwing part of p's precision away in
pbinom.
Using Ambramowitz & Stegun 26.5.2:

-    return pbeta(1 - p, n - x, x + 1, lower_tail, log_p);
+    return pbeta (p, x + 1, n - x, !lower_tail, log_p);

>>>>> "Morten" == Morten Welinder
<terra@gnome.org>
>>>>>     on Mon, 12 Apr 2004 17:20:53 +0200 (CEST) writes:
    Morten> Full_Name: Morten Welinder
    Morten> Version: snapshot
    Morten> OS: 
    Morten> Submission from: (NULL) (65.213.85.129)


    Morten> There is really no good reason for throwing part of
    Morten> p's precision away in pbinom.  Using Ambramowitz &
    Morten> Stegun 26.5.2:

    Morten> -    return pbeta(1 - p, n - x, x + 1, lower_tail, log_p);
    Morten> +    return pbeta (p, x + 1, n - x, !lower_tail, log_p);

Indeed!
Thank you for the fix.

Martin M?chler