0

I was not able to use a self defined (using Module and For loops) piecewise function to fit data. It seems that the problem is that I have to define a pure function. I would like to learn how to define pure functions in the following example using Modules and For loops. Please help.

So let's consider a problem - we have some data, and a self defined piecewise function f[z] that goes exactly through all the data points. LinearModelFit fitting data with f[z] should return value parameter equal to 1.

Disclaimer: it will return a constant instead (basically will not understand what is the f[z] provided and will use a constant to fit the data).

However, in the second part of the following example I will define a function g[z] (which will be equal to f[z]) using Piecewise[] and it will work perfectly (LinearModeFit will return 1).

So, once again, I would like to write f[z] using Module and For loops so that it works in LinearModelFit the way g[z] does.

x = Table[x + Random[], {x, 1, 5}];
y = 2 # + Random[] & /@ x;
data = Transpose@{x, y};

Part 1: self defined function f[z]

fpiecewise[x_, xlist_, ylist_] := Module[{i = 1},
   For[i = 1, i < Length[xlist], i++, If[x <= xlist[[i]], Break[]]];
   Return[ylist[[i]]]
   ];
f[z_] := fpiecewise[z, x, y];

LinearModelFit[data, f[z], z]; 
ListPlot[{{x, % /@ x} // Transpose, data}, Joined -> True]

Part1

Part 2: Piecewise[] defined function g[z]

x1 = Join[{0.}, x];
g[z_] := Piecewise[
  Table[{y[[i]], x1[[i]] < z <= x1[[i + 1]]}, {i, 1, Length[x]}]]
LinearModelFit[data, g[z], z];
ListPlot[{{x, % /@ x} // Transpose, data}, Joined -> True]

Part2

Glorfindel
  • 547
  • 1
  • 8
  • 14
user1541776
  • 451
  • 2
  • 10
  • 1
    Classic mistake fpiecewise[x_?NumberQ, ... – swish May 26 '13 at 22:57
  • @swish, thank you for your help! However, I didn't get it. f[z] was returning expected numeric values when I had fpiecewise[x, ... (without ?NumberQ). I was able to Plot[] it as well. So, what was exactly the problem with using it in Fit[]? Could you, please, refer to a source that discusses this question in more details? – user1541776 May 27 '13 at 09:43

1 Answers1

0

As swish pointed out:

fpiecewise[x_**?NumberQ**, xlist_, ylist_] := Module[{i = 1},
   For[i = 1, i < Length[xlist], i++, If[x <= xlist[[i]], Break[]]];   
   Return[ylist[[i]]]
];

f[z_] := fpiecewise[z, x, y];

LinearModelFit[data, f[z], z];

f[z] defined this way will do the job as successfully as g[z].

user1541776
  • 451
  • 2
  • 10