Re: [dev] Trick for (1-x)^b for x near 1?

2008-07-10 Thread Andrew Douglas Pitonyak
If you have access (I do not) check Journal article by Gary Tietjen; The American Statistician, Vol. 52, 1998 Recursive schemes for calculating common cumulative distributions. Estimating the cumulative distribution function (cdf) for various distributions is at the heart of statistical compu

Re: [dev] Trick for (1-x)^b for x near 1?

2008-07-10 Thread Regina Henschel
Hi Andrew, Andrew Douglas Pitonyak schrieb: Regina Henschel wrote: Hello, I'm working on the beta distribution for Calc. In the case BETADIST(x,1,b,0,1,FALSE()), which is the beta distribution density function, the term b*(1-x)^(b-1) has to be calculated. The domain for x is 0<=x<=1. There

Re: [dev] Trick for (1-x)^b for x near 1?

2008-07-09 Thread Andrew Douglas Pitonyak
Regina Henschel wrote: Hello, I'm working on the beta distribution for Calc. In the case BETADIST(x,1,b,0,1,FALSE()), which is the beta distribution density function, the term b*(1-x)^(b-1) has to be calculated. The domain for x is 0<=x<=1. There is no problem for x<0.99. But you get large

[dev] Trick for (1-x)^b for x near 1?

2008-07-09 Thread Regina Henschel
Hello, I'm working on the beta distribution for Calc. In the case BETADIST(x,1,b,0,1,FALSE()), which is the beta distribution density function, the term b*(1-x)^(b-1) has to be calculated. The domain for x is 0<=x<=1. There is no problem for x<0.99. But you get large relative errors for x nea