User:Regina
From Apache OpenOffice Wiki
- native language
- German
- active in project
- de, qa. sc
- OOo user name
- regina
- special interest
- www.ooowiki.de
Contents
ToDo
Working on functions sides
Functions listed alphabetically, Functions listed by category
Planning Wiki-Pages
FISHER and FISHERINV
Reduced accuracy for r -> 0
Needs log1p for improvement.
1/2*log((1+r)/(1-r)) = 1/2*log(1+2*r/(1-r)) = 1/2* log1p(2r/(1-p))
or
1/2*(log(1+r)-log(1-r)) = 1/2*( log1p(r)-log1p(-r))
Have to test it.
- Test done. Accuracy is the same. But all have additional problem for r -> 1. Seems to be the known problem with 1-r.
Achtung: Fisher(r) = arctanh(r). Letztere hat das gleiche Problem für r -> 0.
Same problem in FISHERINV with (exp(2r)-1)/(exp(2r)+1)
Achtung FISHERINV(r)=tanh(r)
- tanh ist compiler core function in math.h
Sandbox
Test Wiki-markup here