Error in MLE
in [Matlab]
|
From: Leo Ge on 15 Nov 2009 14:00 I defined the normal pdf in M file as: function pdf=betapdf(x,a,b) pdf=1/(2*pi)^0.5/b*exp(-(x-a)^2/2/b^2); And simulate s with normal(0,1) N=2000 p=rand(1,N) s=norminv(p(1:N),0,1) Then mle but just get error: [phat,pci] = mle(s,'pdf',@betapdf,'alpha',.05,'start',[-1,0.5]); ??? Error using ==> stats\private\mlecustom>llf_pdfcdf The PDF function returned NaN or infinite values. Error in ==> fminsearch at 175 fv(:,1) = funfcn(x,varargin{:}); Error in ==> stats\private\mlecustom at 176 [phat,nll,err,output] = ... Error in ==> mle at 219 [phat, pci] = mlecustom(data,varargin{:}); But if I just use [phat,pci] = mle(s,'distribution','normal','alpha',.05); It all works well. I do another practise but also with error: %define beta-df by M-file: function logpdf = betalogpdf(x,a,b) logpdf = (a-1)*log(x)+(b-1)*log(1-x)-betaln(a,b); Then mle by: x = betarnd(1.23,3.45,25,1); phat = mle(x,'logpdf',@betalogpdf,'start',[1,2]) ??? Error using ==> stats\private\mlecustom>checkFunErrs The following error occurred while trying to evaluate the user-supplied logpdf function 'betalogpdf': Error using ==> feval Undefined command/function 'betalogpdf'. Error in ==> stats\private\mlecustom at 156 checkFunErrs('logpdf',logpdfFun,start,uncensData,[],[],logpdfAddArgs); Error in ==> mle at 217 phat = mlecustom(data,varargin{:}); I am confused why I can't specify the pdf by myself. Thank you very much for your kindly help. Leo
From: Leo Ge on 15 Nov 2009 14:46 "Leo Ge" <qwaszxgecl(a)hotmail.com> wrote in message <hdpj42$lml$1(a)fred.mathworks.com>... > I defined the normal pdf in M file as: > function pdf=betapdf(x,a,b) > pdf=1/(2*pi)^0.5/b*exp(-(x-a)^2/2/b^2); > > And simulate s with normal(0,1) > N=2000 > p=rand(1,N) > s=norminv(p(1:N),0,1) > > Then mle but just get error: > [phat,pci] = mle(s,'pdf',@betapdf,'alpha',.05,'start',[-1,0.5]); > ??? Error using ==> stats\private\mlecustom>llf_pdfcdf > The PDF function returned NaN or infinite values. > Error in ==> fminsearch at 175 > fv(:,1) = funfcn(x,varargin{:}); > Error in ==> stats\private\mlecustom at 176 > [phat,nll,err,output] = ... > Error in ==> mle at 219 > [phat, pci] = mlecustom(data,varargin{:}); > > But if I just use > [phat,pci] = mle(s,'distribution','normal','alpha',.05); > It all works well. > > I do another practise but also with error: > %define beta-df by M-file: > function logpdf = betalogpdf(x,a,b) > logpdf = (a-1)*log(x)+(b-1)*log(1-x)-betaln(a,b); > Then mle by: > x = betarnd(1.23,3.45,25,1); > phat = mle(x,'logpdf',@betalogpdf,'start',[1,2]) > ??? Error using ==> stats\private\mlecustom>checkFunErrs > The following error occurred while trying to evaluate > the user-supplied logpdf function 'betalogpdf': > Error using ==> feval > Undefined command/function 'betalogpdf'. > Error in ==> stats\private\mlecustom at 156 > checkFunErrs('logpdf',logpdfFun,start,uncensData,[],[],logpdfAddArgs); > Error in ==> mle at 217 > phat = mlecustom(data,varargin{:}); > > I am confused why I can't specify the pdf by myself. Thank you very much for your kindly help. > > Leo I find a problem that I don't save the m-file with the same name of the function.But after revising it, the first example still have error: ??? Error using ==> stats\private\mlecustom>checkFunErrs The following error occurred while trying to evaluate the user-supplied logpdf function 'normlogpdf': Error using ==> mpower Matrix must be square. Error in ==> stats\private\mlecustom at 156 checkFunErrs('logpdf',logpdfFun,start,uncensData,[],[],logpdfAddArgs); Error in ==> mle at 219 [phat, pci] = mlecustom(data,varargin{:}); I don't know what's wrong. I am a newer to Matlab, and thank you for your kindly help.
From: Tom Lane on 16 Nov 2009 10:13 >> pdf=1/(2*pi)^0.5/b*exp(-(x-a)^2/2/b^2); .... > Error using ==> mpower > Matrix must be square. Leo, the expression you wrote is fine as long as everything is a scalar. But if x is a vector, (x-a)^2 requests that you square the entire array, using matrix multiplication. You can use (x-a).^2 if your intent is to square each element of the array. -- Tom
From: Leo Ge on 16 Nov 2009 17:08 Yeah, it works, many thx!
|
Pages: 1 Prev: How to plot values while using Matlab optimization toolbox Next: Simulink OPC configuration |