Dear Sir I am new user of R. I am interested in modeling hydrological extreme events. I found MSClaio2008 very interesting function. In this function four criterions for choosing distributions. Can we call these criterions as model selection techniques or goodness of fit techniques or both? Because goodness of fit techniques are usually performed after modle selection. Can I found chi-square, kolmogrov-sminov and cramer-von mises tests for testing goodness of fit for proposed distributions? Please help _________________________________________________________________ Show them the way! Add maps and directions to your party invites. [[alternative HTML version deleted]]
> I am interested in modeling hydrological extreme events. I found > MSClaio2008 very interesting function. In this function four > criterions for choosing distributions. Can we call these criterions > as model selection techniques or goodness of fit techniques or both? > Because goodness of fit techniques are usually performed after modle > selection.What is MSClaio2008? I don't quite understand this bit. If you provide code examples, then questions are often easier to answer.> Can I found chi-square, kolmogrov-sminov and cramer-von mises tests > for testing goodness of fit for proposed distributions?Yes, you can do these things in R. The functions you want are: chisq.test for the chi-square test ks.test for the kolmogorov-smirnoff test. You can view the help pages for these functions by typing a question mark before the function name, e.g. ?chisq.test. For the Cramer-von Mises tests, I didn't know, so I typed RSiteSearch("cramer von mises test"), from which there are several suggestions. In particular, look at the CvM2SL1Test and CvM2SL2Test packages. Regards, Richie. Mathematical Sciences Unit HSL ------------------------------------------------------------------------ ATTENTION: This message contains privileged and confidential inform...{{dropped:20}}
I want to fit a kernel density to a data-set of size 1000 and each data point being of dimension 5. The default density function does'nt work for me. Is there any other function available which can do the job efficiently.........? I don't have access to a very fast computer.
(copy of earlier reply which was not addressed to r-help. Also added informative Subject:) "yes ......" is not responsive to the question of how you propose to display or examine such a mathematical object. My signature was perhaps a lame (certainly an ineffective) effort at getting you to acknowledge that there might, just might, be some conceptual difficulties in "viewing" a 5+1 = 6 dimensional object. Can you point to any examples of such a procedure being applied in a manner that you find helpful? It is certainly feasible to calculate distances in 5-space, but the next step, displaying the "concentration" of those distances as a function of the coordinates, would be the "tough nut". (1000 data points is not a problem. That amount of data should easily fit in any device that can run R.) -- David On Jun 4, 2009, at 12:06 AM, arijit kumar debnath wrote:> yes...... > I want a density estimator for 5 dimensional data.I want it to be > efficient since I don't have access to a fast computer.My data set > is quite large(about 1000 data points) and I have only a pentium 4 > 1.7 Ghz (512 Mb RAM)computer. > > On Thu, Jun 4, 2009 at 7:25 AM, David Winsemius <dwinsemius@comcast.net > > wrote: > We can get you a kde2d but you really want a kde5d? What sort of > display are you proposing? > > -- > David "Earthling" Winsemius > > On Jun 3, 2009, at 9:26 PM, arijit kumar debnath wrote: > > I want to fit a kernel density to a data-set of size 1000 and each > data point being of dimension 5. The default density function does'nt > work for me. Is there any other function available which can do the > job efficiently.........? > I don't have access to a very fast computer.David Winsemius, MD Heritage Laboratories West Hartford, CT [[alternative HTML version deleted]]