Re: Interpolation undocumented error

*To*: mathgroup at smc.vnet.net*Subject*: [mg113664] Re: Interpolation undocumented error*From*: "Ingolf Dahl" <ingolf.dahl at telia.com>*Date*: Sat, 6 Nov 2010 05:01:59 -0500 (EST)

Maybe you have some points with identical or very close coordinates? Then you might get this kind of error for the RBF interpolation. It would be easy to put a filter in the interpolation routine before the interpolation, but I did not find it attractive to tamper with the input data behind the back of the user. It is thus the responsibility of the user to remove such doublets, and I think it should be that way. Another warning: RBF interpolation can be very good for "mathematical data", e.g. solutions of differential equations, but can give big oscillations for noisy experimental data. Best regards Ingolf Dahl > -----Original Message----- > From: Hyatt, John S [mailto:jshyatt at gatech.edu] > Sent: den 5 november 2010 11:13 > To: mathgroup at smc.vnet.net > Subject: [mg113630] Re: Interpolation undocumented error > > Hi all, > > I have tried both the Obtuse and Imtek packages suggested. They are installed and working > correctly. However, I cannot get a meaningful interpolation function to my data. Using > Imtek's imsUnstructuredInterpolation, I get the following error message: > > LinearSolve::luc: Result for LinearSolve of badly conditioned matrix {{-3.1356*10^21,- > 2.01921*10^21,-1.27651*10^21,-7.89872*10^20,-4.77313*10^20,- > 2.81064*10^20,<<40>>,-5.56912*10^18,-5.56912*10^18,-2.43624*10^18,- > 2.43624*10^18,<<315>>},<<49>>,<<315>>} may contain significant numerical errors. >> > > And for Obtuse, using the RBF method for Interpolation, I get: > > LinearSolve::luc: Result for LinearSolve of badly conditioned matrix {<<1>>} may contain > significant numerical errors. >> > > Both of these messages are similar. However, I don't know how to make sense of this error > either: first, what is <<...>>, and second, what is the message trying to say? The data is > single-valued and obviously lies on a smooth surface - I can eyeball what a given value > should be, but I want a specific interpolation function so I can make it part of my analysis > routine. Does anyone recognize this message, and for those of you familiar with Obtuse and > Imtek, has this problem cropped up before? As far as I can tell I'm setting up my data and > interpolation in the same format as in the examples in both packages' documentation. I am > making sure that the third element of each data point is the "z" value, and that the "x" and > "y" designations are consistent within all the different data sets I'm trying to interpolate. > I've tried different orders of fitting and different fit types, and each time, while I do get an > interpolation surface, it's complete nonsense (i.e. nowhere near t! > he data). > > Thanks in advance, > John H > > ----- Original Message ----- > From: "Ingolf Dahl" <Ingolf.Dahl at physics.gu.se> > To: mathgroup at smc.vnet.net > Sent: Thursday, October 21, 2010 7:03:01 AM > Subject: [mg113630] [mg113269] Re: Interpolation undocumented error > > You might try my "Obtuse" package, downloadable from http://www.familydahl. > se/mathematica/ . > This package was inspired by exactly this error message, and it contains five different > interpolation methods, defined as "methods" to the Interpolation function. Which method to > choose depends strongly on the character of the data, so you probably need to experiment a > little to see what suits your case. > > Best regards > > Ingolf Dahl > Ingolf.dahl at telia.com > > -----Ursprungligt meddelande----- > Fr==E5n: Hyatt, John S [mailto:jshyatt at gatech.edu] > Skickat: den 20 oktober 2010 10:06 > Till: mathgroup at smc.vnet.net > ==C4mne: [mg113232] Interpolation undocumented error > > I am trying to use Interpolation to get an interpolation function of multidimensional data > from light scattering. I specifically want to be able to calculate the value of the function at > any point within certain parameters. I have entered the data points as usual into > Interpolation, but receive an error message: > > "Interpolation::indim: The coordinates do not lie on a structured tensor product grid." > > This error is undocumented in the Mathematica Help as far as I can tell, an > d I don't know what the error means (i.e. what is a structured tensor produ= > == > ct grid). After searching the web for other people with the same issue, I = > == > found only one thread on Wolfram's archives where the Imtek Mathematica Sup= > == > plement was suggested as a fix and seemed to help the person. However, the= > == > IMS seems to be intended for use with an earlier version of Mathematica (I= > == > have v7.0) and the installation instructions in any case aren't applicable= > == > to this version. Is there a fix for this in Mathematica, or does anyone k= > == > now of any way to get the interpolation function I am looking for? Would u= > == > ninstalling Mathematica and installing version 6 followed by the IMS be the= > == > only way to fix this? > > Thanks in advance for any help, > John H