0

This is my first time using NonLinearModelFit and I have run into an issue that I can't find an explanation for. The fitted model contains a <<19>> in it but when I look at the estimates for the parameters using "ParameterConfidenceIntervalTable" a value has been calculated for the parameter whose place has been taken by <<19>>. I found that using "BestFit" will give me the proper fitted model but I want to know why <<19>> was there in the first place and where it came from. I couldn't find an explanation for this specific issue elsewhere.

Here is a sample of my code and the fitted model generated before using "BestFit"

data = {{1.0, 1.0, 0.051}, {1.0, 2.0, 0.072}, {2.0, 3.0, 0.16}, {0.5, 
    2.0, 0.033}, {4.0, 5.5, 0.49}, {1.5, 4.0, 0.14}, {6.4, 2.0, 
    0.43}, {6.0, 4.0, 0.62}};
soln = NonlinearModelFit[data, k*cA^n*cB^m , {k, n, m}, {cA, cB}, 
  MaxIterations -> 200]
soln["ParameterConfidenceIntervalTable"]

FittedModel[ 0.04106 cA^<<19>> cB^0.607558 ]

According to the confidence interval table, the estimate for n (<<19>>) is 1.04206. Using "BestFit" I get this result for the fitted model:

soln["BestFit"]
0.0410613 cA^1.04206 cB^0.607558

Why is that? Is there a way to get that solution without using "BestFit"?

baumannr
  • 787
  • 3
  • 10
  • Try Normal@soln. If I understand correctly your concern is the Skeleton. – Kuba Aug 13 '13 at 18:25
  • << ...>> is used to print results in shortened form, or to hide some detail considered to be less necessary. The results returned by NonlinearModelFit are an example of this. You get a object that can be 'questioned' for various interesting things that it knows of itself. You do this by passing arguments to the object (like "BestFit"). What you see as FittedModel[ 0.04106 cA^<<19>> cB^0.607558 ] is just a visual representation of this object, a tag, or a label and not really meant to be the answer. The 19 in the label is an indication of how much data (in the label) has been hidden. – Sjoerd C. de Vries Aug 13 '13 at 18:30
  • So, using soln["BestFit"] is really the intended way to get your fit, although you also could use Normal. You could also attach ["BestFit"] to the NonlinearModelFit function call like this: NonlinearModelFit[...]["BestFit"] – Sjoerd C. de Vries Aug 13 '13 at 18:36
  • @SjoerdC.deVries I think that if you are not voting to close as "simple mistake" you should post your comments as an answer :) – Kuba Aug 13 '13 at 19:09
  • @kuba I can't, you already closed it ;-) – Sjoerd C. de Vries Aug 14 '13 at 06:30
  • @SjoerdC.deVries oh, indeed, but you had 3 hours to do so :) – Kuba Aug 14 '13 at 06:31

0 Answers0