Nan Zhao <nzhao <at> student.ethz.ch> writes: > > Thank you Dennis for your explanations! > > The results you found are the same as mine. with first an infinity result, > followed by NaN. It seems that, when the number becomes too small, R must > round it up to 0. Hence I was wondering if there might a way to increase the > number of decimals for extremal computations. I have tried to use "double" > variables, but this didn't have a bigger success. > > There is no doubt about the function and it orignates from a Kendall's plot > in measuring dependencies. > > Thank you in advance, > > Best Regards, > > Nan >
See the thread "Problem with the precision of numbers" in January this year. For example in http://finzi.psych.upenn.edu/Rhelp10/2010-January/225640.html you will also find a hint how to compute binomial coefficients with 'Rmpfr'. Integration will be a problem here that you may have to replace with your own method for integration, such as Simpson's adaptive formula. Regards, Hans Werner ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.