Mathematica 9 is now available
Student Support Forum
-----
Student Support Forum: 'NonlinearFit' topicStudent Support Forum > General > Archives > "NonlinearFit"

Next Comment >Help | Reply To Topic
Author Comment/Response
Nathan
07/09/04 6:36pm

I have this problem when using NonlinearFit with the following model and data:

data = {{50, 3225.0567}, {100, 6108.877}, {150, 8531.3910}, {200, 10708.2915}, {250, 12556.8653}, {300, 14154.2851}, {350, 15422.3550`}, {400, 16834.3916}, {450, 17906.3075}, {500, 18767.5784}, {550, 19418.7275}, {600, 20223.4710}, {650, 20955.6299}, {700, 21396.0879}};

NonlinearRegress[data, 17970*(ku/kd)*(1 - \[ExponentialE]^(-kd*x)), {x}, {ku, kd}]

this code returns a warning:
NonlinearRegress::"rank": "Warning: Parameter solution locus has \
dimensionality of only 1 at the parameter estimate, less than full \
dimensionality 2. Confidence region curvature given by \
1/Sqrt[Quantile[FRatioDistribution[1, 12], 0.95`]]"

and procedes to give me terribly bad estimates

BestFitParameters -> {ku -> 328.405, kd -> 400.601},

where, actually, ku=.00386 and kd=.00287

I'm not sure what's the matter here. SAS will do this nonlinear fit just fine (estimating .00388,.00280) but I'd like to do it in Mathematica because they rest of my project is in Mathematica. Anybody knows what the problem is here?


URL: ,

Subject (listing for 'NonlinearFit')
Author Date Posted
NonlinearFit Nathan 07/09/04 6:36pm
Re: NonlinearFit Henry Lamb 07/10/04 9:00pm
Next Comment >Help | Reply To Topic