Apache OpenOffice (AOO) Bugzilla – Issue 26836
Precision Problem in OOo 1.1.0 Calc (NORMINV)
Last modified: 2013-08-07 15:13:53 UTC
I use OpenOffice.org 1.1.0 in mathematics in school, and I have a heavy problem: NORMINV(0,975;0;1) Should output something like 1.959962, but the result in OOo is 1,960395!! Validated the value with some Calculators like TI-92 and TI-200 Voyage and excel...
Created attachment 13988 [details] Document containing the NORMINV function
Hi Eike, please have a look at this. I've attached a document with this function. Frank
Yes! Exactly this problem! I proofed the values from the sheety by my TI Calculator and the HP Calculator of a friend... (I'm sorry that I forgot to post my spreadsheet as example as you, fst ;-)
Another manifestation of issue 18704. Adding needhelp keyword and targeting to OOoLater right now for time reasons. Implementation is in sc/source/core/tool/interpr3.cxx ScInterpreter::ScNormInv() respectively the gaussinv() method used therein.
Ha Haa!!! Okay, I got it... Was not easy, but I played arround with it, found out that t steps are not deph enough and so on... But after a small look at the way Gnumeric implements it - I found out they use AS241, I tried too... It's a little bit different implement, but it works perfectly. Should solve all problems with gaussinv! I hope it is usefull for you and you can integrate it into 1.1.1 ;-) tried it with all statistical functions i know using gaussinv with nearly perfect results, but PLEASE test it too! Nobody is perfect, maybe I did bullshit somewhere without see'ing it ;-) The workarround is postet as simple C-File! Have Fun! Martin
Created attachment 14012 [details] New gaussinv-implementation for OOo Calc (gaussinv.c)
We can't use any code without the JCA (Joint Copyright Assignment, see http://www.openoffice.org/contributing.html).
psiplus, are you willing to submit a JCA for the code?
Hi I just sent it by fax :-) ... Sorry, I forgot about it... I hope the pice of code can now be added, and maybe added to OOo2 ;-) ... ... I work with it for more than one now without problems... Have Fun! Martin
I'm taking care of this.
I put this into the CWS "dr42" (interpr3.cxx 1.14.2.1), so it will get into OOo 2.0.1. Thanks a lot for the patch, Martin. For testing: This affects functions NORMINV, NORMSINV, LOGINV and CONFIDENCE. Note: The related issue 32833 (to use iteration instead of the approximation formula) is still open, but might actually be obsoleted now.
back to QA for verification re-open issue and reassign to fst@openoffice.org
reassign to fst@openoffice.org
reset resolution to FIXED
found fixed on cws dr42 using Linux, Solaris and Windows build
Found integrated on master m142 using Linux, Solaris and Windows Build
Created attachment 73205