RE: Nonparametric kernel density estimators
- To: mathgroup at smc.vnet.net
- Subject: [mg38526] RE: Nonparametric kernel density estimators
- From: "DrBob" <drbob at bigfoot.com>
- Date: Fri, 20 Dec 2002 23:40:57 -0500 (EST)
- Reply-to: <majort at cox-internet.com>
- Sender: owner-wri-mathgroup at wolfram.com
Yes, we can do it ourselves, but MathStatica isn't of much help in the process, and it can't be loaded after other packages, so the net benefit may be negative. As for Bandwidth, how does it decide what is optimal? Can't I do that in Mathematica, too? When I try to use MathStatica, I first have to eliminate everything in my Init.m file (and change it back again later), which is a pain. Once I get into it and start dealing with the quirky interface, the unnecessary splash screen and web-site palette, etc. --- it just never seems worthwhile. Bobby -----Original Message----- From: Bill Rowe [mailto:listuser at earthlink.net] To: mathgroup at smc.vnet.net Subject: [mg38526] Re: Nonparametric kernel density estimators On 12/20/02 at 6:50 AM, drbob at bigfoot.com wrote: >As best I can tell, MathStatica graphs kernel density estimates, but >gives you no other access to them. Yes, MathStatica provides no functions for directly manipulating non-parametric kernel density estimates except for plotting them. But I do not see this as being a significant limitation. If you need more than a plot, you could either extract the points from the plot MathStatica produces and fit an interpolating function to it or you could use ListConvolve with the desired kernel to compute the values. As I see it the biggest problem with doing non-parametric kernel density estimation is determing the optimum bandwidth to use. MathStatica provides a function appropriately named Bandwidth to compute this from the data and a given kernel