Putting a matrix inside function

Question

I want to fit a function to some data and inside the function there's a list

here's my data

Ei = {{1, -0.00006636511699298353`}, {2, -0.00007582534298933297`}, \
{3, -0.00008658039298836684`}, {4, -0.00009882137899808185`}, {5, \
-0.0001127647090015671`}, {6, -0.00012865092799074773`}, {7, \
-0.00014674610198994742`}, {8, -0.00016733831999715676`}};

the list

racc = {6.0, 5.9, 5.8, 5.7, 5.6, 5.5, 5.4, 5.3, 5.2, 5.1};

and my function

U[r_] := ecc (1 - (racc[[r]] - Exp[-Acc (1 - rscc)])^2);

But mathematica shows my function as

ecc (1 - (-E^(-Acc (1 - rscc)) + {6., 5.9, 5.8, 5.7, 5.6, 5.5, 5.4, 
   5.3, 5.2, 5.1}[[r]])^2)

Though in someway it can understand it and give the result for function

U[3] /. {ehh -> 1, Ahh -> 1, rshh -> 1, ech -> 1, Achrsch -> 1, 
ecc -> 1, Acc -> 1, rscc -> 1};

-22.04

And it's able to fit it, although an giving an error

sol = FindFit[Ei, U[r], {ecc, Acc, rscc}, r]

Part::pspec: Part specification r is neither an integer nor a list of integers. >>    
{ecc -> 3.45136*10^-6, Acc -> -21.5489, rscc -> 2.67028}

It can not Plot it

P2 = Plot[U[r] /. sol, {r, 1, 8}];
Part::pspec: Part specification 1.000143` is neither an integer nor a list of integers. >>
Part::pspec: Part specification 1.1430001428571428` is neither an integer nor a list of integers. >>
Part::pspec: Part specification 1.2858572857142856` is neither an integer nor a list of integers. >>
General::stop: Further output of Part::pspec will be suppressed during this calculation. >>

Does anybody have an idea how to put a matrix or list inside a function?

score 2 · Accepted Answer · answered May 08 '13 at 07:56

2

Try FromDigits[]

U[r_] := ecc (1 - (FromDigits[racc, r] - Exp[-Acc (1 - rscc)])^2);

answered May 08 '13 at 07:56

Raymond Ghaffarian Shirazi

952
6
15

score 2 · Answer 2 · answered May 08 '13 at 08:16

2

I think the issue is that your function U is defined only for integer r, whereas FindFit would want to sample the function at intermediate points as well. In this case, you can build the least squares by hand. Additionally, your choice seems to be a poor fit; the vector racc decreases with its index, which makes your U increase so you won't get a good fit. Just for fun, I tried another fitting function with racc Reversed.

bigU[r_] := ecc (1 - (racc[[r]] - Exp[-Acc (1 - rscc)])^2)
bigU2[r_] := ecc (1 - (Reverse[racc][[r]] - ex)^2)

sol = NMinimize[Total[(bigU[#[[1]]] - #[[2]])^2 & /@ Ei], {ecc, Acc, rscc}];
sol2 = NMinimize[Total[(bigU2[#[[1]]] - #[[2]])^2 & /@ Ei], {ecc, ex}];

ListPlot[{Ei, {#, bigU[#] /. sol[[2]]} & /@  Ei[[All, 1]], 
          {#, bigU2[#] /. sol2[[2]]} & /@ Ei[[All, 1]]}, Joined -> {False, True, True}]

plot

answered May 08 '13 at 08:16

b.gates.you.know.what

20,103
2
43
84

is there a way to stop this error "Part::pspec: Part specification r is neither an integer nor a list of integers. >>"
I can not select any arrays or etc in find fit
– Sarah Aria May 08 '13 at 08:33
You can make U a function of a generic real variable. – b.gates.you.know.what May 08 '13 at 08:41
To clarify one point: FindFit (and NonlinearModelFit, which is based on it) do not sample at any other points than those given explicitly. That is what makes e.g. this technique possible. – Oleksandr R. May 11 '13 at 02:18
@OleksandrR. Thanks, I didn't know that and I had missed your answer. – b.gates.you.know.what May 11 '13 at 10:23

Putting a matrix inside function

2 Answers2