Line Equations from Interpolated Data

Question

I have an interpolated line from a series of measured data points. Is there a way I can get Mathematica to output an approximate equation for the interpolated line?

For example: From the interpolation documentation is this graph:

The first part of the line, 0 < x < 4 seems to approximate to a polynomial 1 line - Could Mathematica output the 'best fit' for this line, and something like an R^2 value for how close the fit is?

The same for 4 < x < 6, however this appear to approximate to an polynomial 2, so the output would be something of the form y = ax^2 + bx + c.

Or for the whole domain?

Basically i would like to take a data set > plot an interpolated line on a graph > then get mathematica to output the approximate line equation and output details on how good the fit is.

Thanks in advance guys!

Depends on the method of interpolation. Here's a possible duplicate: https://mathematica.stackexchange.com/questions/59944/extracting-the-function-from-interpolatingfunction-object/59963#59963 -- If it doesn't work for you, a complete example will be needed. — Michael E2, Dec 05 '17 at 02:04

score 4 · Answer 1 · answered Dec 05 '17 at 03:35

data = {{1, 1}, {2, 2}, {3, 3}, {4, 5}, {5, 8}, {6, 5}};

if = Interpolation[data];

The InterpolatingFunction passes thru the data points

(if /@ data[[All, 1]]) == data[[All, 2]]

(* True *)

Plot[if[x], {x, 1, 6}, Epilog -> {Red, AbsolutePointSize[5], Point[data]}]

The InterpolatingPolynomial will also pass thru the data points.

ip[x_] = InterpolatingPolynomial[data, x] // Expand

(* 4 - (527*x)/60 + (211*x^2)/24 - 
   (11*x^3)/3 + (17*x^4)/24 - x^5/20 *)

(ip /@ data[[All, 1]]) == data[[All, 2]]

(* True *)

Plot[ip[x], {x, 1, 6}, Epilog -> {Red, AbsolutePointSize[5], Point[data]}]

While the InterpolatingFunction and InterpolatingPolynomial appear similar, they are different. For example, comparing their maximum values

NMaximize[{#, 1 <= x <= 6}, x] & /@ {if[x], ip[x]}

(* {{8.10605, {x -> 5.17631}}, {8.33528, {x -> 5.2972}}} *)

You can also use FindFormula. Since FindFormula uses RandomSeeding it can give varying results. However, it can produce a numeric form of the InterpolatingPolynomial.

ff[x_] = FindFormula[data, x] // Rationalize[#, 10^-7] &

(* 4 - (527*x)/60 + (211*x^2)/24 - 
   (11*x^3)/3 + (17*x^4)/24 - x^5/20 *)

ff[x] == ip[x]

(* True *)

For your requested model of line joined with a quadratic, use Piecewise

model[a_, b_, c_, d_, e_, x_] = 
  Piecewise[{{a x, 1 <= x < e}, {b x^2 + c x + e (a - c - b e), 
     e <= x <= 6}}];

nlm[x_] = NonlinearModelFit[data, model[a, b, c, d, e, x], {a, b, c, d, e}, 
    x]["BestFit"] /. z_?NumericQ :> RootApproximant[z]

The nlm also passes thru the data points

(nlm /@ data[[All, 1]]) == data[[All, 2]]

(* True *)

Plot[nlm[x], {x, 1, 6}, Epilog -> {Red, AbsolutePointSize[5], Point[data]}]

Line Equations from Interpolated Data

1 Answers1