how to fit distribution if pdf known
in [Matlab]
|
From: HollyandIvy He on 12 Apr 2010 09:48 I'm trying to fit a pdf into distribution, for instance, normal distribution, but I found normfit works in a different way. [URL=http://www.slide.com/s/iakWwZLe7z_dVpF_h0DraIPxpdRCZszI?referrer=hlnk][IMG]http://widget.slide.com/rdr/1/1/4/S/2d00000020d1a7cf/1/93/oMUnwuw65j9nUrzsANd8tAS1adOK-SUn.jpg[/IMG][/URL] this is what I want to fit, I know it is not a standard distribution, I just want to know if there is direct command in Matlab can fit into a form like N~(1104,800) rather than a N~(0.0213 0.0155),later is the output of using 'normfit'
From: Peter Perkins on 12 Apr 2010 10:12 On 4/12/2010 9:48 AM, HollyandIvy He wrote: > I'm trying to fit a pdf into distribution, for instance, normal > distribution, but I found normfit works in a different way. > [URL=http://www.slide.com/s/iakWwZLe7z_dVpF_h0DraIPxpdRCZszI?referrer=hlnk][IMG]http://widget.slide.com/rdr/1/1/4/S/2d00000020d1a7cf/1/93/oMUnwuw65j9nUrzsANd8tAS1adOK-SUn.jpg[/IMG][/URL] > > this is what I want to fit, I know it is not a standard distribution, I > just want to know if there is direct command in Matlab can fit into a > form like N~(1104,800) rather than a N~(0.0213 0.0155),later is the > output of using 'normfit' I can't tell what you are asking for, but here are a couple of possibilities: The data in the image you link to appears to be multimodel. You might want to fit a Gaussian mixure distribution, which you can do with GMDISTRIBUTION. NORMFIT or FITDIST estimate parameters of a normal distribution, given data. You can also create a normal probability distribution using known parameter values like this: >> dist = ProbDistUnivParam('Normal',[1104,sqrt(800)]) dist = normal distribution mu = 1104 sigma = 28.2843 Typing " help ProbDistUnivParam" will more or less show you what you can do with it. Hope this helps.
From: Tom Lane on 12 Apr 2010 10:14 > I'm trying to fit a pdf into distribution, for instance, normal > distribution, but I found normfit works in a different way. > [URL=http://www.slide.com/s/iakWwZLe7z_dVpF_h0DraIPxpdRCZszI?referrer=hlnk][IMG]http://widget.slide.com/rdr/1/1/4/S/2d00000020d1a7cf/1/93/oMUnwuw65j9nUrzsANd8tAS1adOK-SUn.jpg[/IMG][/URL] > this is what I want to fit, I know it is not a standard distribution, I > just want to know if there is direct command in Matlab can fit into a form > like N~(1104,800) rather than a N~(0.0213 0.0155),later is the output of > using 'normfit' Try "help mle" for information about fitting a distribution when you want to supply the form of the pdf. If you have values from a pdf, and you want to fit a density-shaped curve to them, then you might want to take a look at this demo: http://www.mathworks.com/products/statistics/demos.html?file=/products/demos/shipping/stats/cfitdfitdemo.html -- Tom
From: HollyandIvy He on 12 Apr 2010 12:08 > I can't tell what you are asking for, but here are a couple of possibilities: > > The data in the image you link to appears to be multimodel. You might want to fit a Gaussian mixure distribution, which you can do with GMDISTRIBUTION. > > NORMFIT or FITDIST estimate parameters of a normal distribution, given data. You can also create a normal probability distribution using known parameter values like this: > > >> dist = ProbDistUnivParam('Normal',[1104,sqrt(800)]) > dist = > normal distribution > mu = 1104 > sigma = 28.2843 > > Typing " help ProbDistUnivParam" will more or less show you what you can do with it. Hope this helps. That is my problem that I can't get mean and variance, in fact I need these parameters for random samples generation from the distribution.
From: HollyandIvy He on 12 Apr 2010 12:17
"Tom Lane" <tlane(a)mathworks.com> wrote in message <hpv9s9$nar$1(a)fred.mathworks.com>... > > I'm trying to fit a pdf into distribution, for instance, normal > > distribution, but I found normfit works in a different way. > > [URL=http://www.slide.com/s/iakWwZLe7z_dVpF_h0DraIPxpdRCZszI?referrer=hlnk][IMG]http://widget.slide.com/rdr/1/1/4/S/2d00000020d1a7cf/1/93/oMUnwuw65j9nUrzsANd8tAS1adOK-SUn.jpg[/IMG][/URL] > > this is what I want to fit, I know it is not a standard distribution, I > > just want to know if there is direct command in Matlab can fit into a form > > like N~(1104,800) rather than a N~(0.0213 0.0155),later is the output of > > using 'normfit' > > Try "help mle" for information about fitting a distribution when you want to > supply the form of the pdf. > > If you have values from a pdf, and you want to fit a density-shaped curve to > them, then you might want to take a look at this demo: > > http://www.mathworks.com/products/statistics/demos.html?file=/products/demos/shipping/stats/cfitdfitdemo.html > > -- Tom > That's the problem, Idon't have sample values but thispure pdf function, mle is fitting sample values. It seems that the link you given is also concentrted on sample fitting while I need find out the parameter define a pdf. if using mle, i'll get N~(0.0213 0.0155), but from the curve obviously the mean is around 800 and 1200 seperately. |