I haven't quite worked it out for the Wiebull but that will certainly (down-) bias your variance estimates for, e.g., the normal.
I think the right thing to do is to use the correct distribution rather than forcing an incorrect distribution to fit incorrectly. Michael On Feb 21, 2012, at 1:10 PM, Vanúcia Schumacher<vanucia-schumac...@hotmail.com> wrote: > > Hi, > then, If I replace the midpoint 0.0025 average values equal to 0there won't > be prejudice in the results?Would then be the best thing to do? > > regards, > > > >> Subject: Re: [R] help error: In dweibull(x, shape, scale, log) : NaNs >> produzidos >> From: pda...@gmail.com >> Date: Tue, 21 Feb 2012 18:55:51 +0100 >> To: vanucia-schumac...@hotmail.com >> >> >> On Feb 21, 2012, at 18:34 , Van�cia Schumacher wrote: >> >>> Thank's, >>> >>> but then I could simply round off my data which are equal to zero to >>> 0.00001? >> >> Something like that. I'd maybe try something a little more carefully >> considered. E.g. if your data are rounded to two decimal figures, the >> "zeros" represent values of x with >> >> 0 < x < 0.005 >> >> since anything bigger would be rounded to 0.01. So replace with the interval >> midpoint of 0.0025. >> >> (One can do better by modifying the likelihood, but then you essentially >> have to rewrite fitdistr().) >> >>> >>> Att, >>> >>> Van�cia Schumacher >>> Curso de gradua��o em meteorologia - UFPEL >>> Bolsista do Programa de Educa��o Tutorial - PET >>> >>> >>>> Subject: Re: [R] help error: In dweibull(x, shape, scale, log) : NaNs >>>> produzidos >>>> From: pda...@gmail.com >>>> Date: Tue, 21 Feb 2012 17:43:07 +0100 >>>> CC: vanucia-schumac...@hotmail.com; r-help@r-project.org >>>> To: dwinsem...@comcast.net >>>> >>>> >>>> On Feb 21, 2012, at 16:36 , David Winsemius wrote: >>>> >>>>> It suggests you managed to send negative or infinite numbers to dweibull, >>>>> a distribution which only supports values when given positive, finite >>>>> numbers. Look at your data more closely. >>>> >>>> Zeros are a common cause, too. For a < 1 (the shape parameter), the >>>> Weibull density has a singularity at zero, and fitdistr() is not tolerant >>>> of observations rounded to zero in such cases. >>>> >>>> -- >>>> Peter Dalgaard, Professor >>>> Center for Statistics, Copenhagen Business School >>>> Solbjerg Plads 3, 2000 Frederiksberg, Denmark >>>> Phone: (+45)38153501 >>>> Email: pd....@cbs.dk Priv: pda...@gmail.com >>>> >> >> -- >> Peter Dalgaard, Professor >> Center for Statistics, Copenhagen Business School >> Solbjerg Plads 3, 2000 Frederiksberg, Denmark >> Phone: (+45)38153501 >> Email: pd....@cbs.dk Priv: pda...@gmail.com >> > > [[alternative HTML version deleted]] > > ______________________________________________ > 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. ______________________________________________ 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.
