Fitting with a Piecewise Linear Function


In this tutorial we will show you how to define a piecewise fitting function consisting of two linear segments, perform a fit of the data using this fitting function, and calculate the intersection location for two linear segments from the fitting result.

Minimum Origin Version Required: Origin 8.6 SR0

What you will learn

This tutorial will show you how to:

  • Define a piecewise (conditional) fitting function.
  • Auto initialize parameters.
  • Calculate the intersection location of the piecewise fit lines.

Example and Steps

Import Data

  1. Open a new workbook.
  2. Click the Import Single ASCII button Import Single ASCII.png to bring up the Open dialog. Browse to \Samples\Curve Fitting folder and select the file Step01.dat.
  3. Right click on the Sensor E x column (column J), and select Set As: X from the context menu. Highlight Sensor E y column, and select Plot: Symbol: Scatter from Origin menu. The graph should look like:
Fit PWL G1.png

Define Fitting Function

From the above graph, the curve consists of two segments of lines. It can be fitted with a piecewise linear function. The function can be expressed as:

y =
\frac{y_1(x_3-x)+y_3(x-x_1)}{x_3-x_1}, & \mbox{if } x<x_3 \\
\frac{y_3(x_2-x)+y_2(x-x_3)}{x_2-x_3}, & \mbox{if } x \ge x_3

where x1 and x2 are x values of the curve's endpoints and they are fixed during fitting, x3 is the x value at the intersection of two segments, and y1, y2, y3 are y values at \ x_i, \ i=1, 2, 3 respectively.

The fitting function can be defined using the Fitting Function Builder tool.

  1. Select Tools: Fitting Function Builder from Origin menu.
  2. In the Fitting Function Builder dialog's Goal page, click Next button.
  3. In the Name and Type page, select User Defined from Select or create a Category drop-down list, type pwl2s in the Function Name field, and select Origin C in Function Type group. And click Next button.
  4. In the Variables and Parameters page, type x1,y1,x2,y2,x3,y3 in the Parameters field. Click Next button.
  5. In the Origin C Fitting Function page, click the User-Defined Fitting Functions-2.png button on the right of the Function Body edit box and define the fitting function in Code Builder as follows.
       if( x < x3 )
          y = (y1*(x3-x)+y3*(x-x1))/(x3-x1);
          y = (y3*(x2-x)+y2*(x-x3))/(x2-x3);

    Click Compile button to compile the function body. Then click Return to Dialog button. Click Next button.

  6. In the Parameter Initialization Code page, click the User-Defined Fitting Functions-2.png button on the right of the Initialization Code edit box and initialize the fitting parameters in Code Builder as follows.
       int n1, n2, n3;
       x_data.GetMinMax( x1, x2, &n1, &n2 );
       x3 = x1 + (x2 - x1)/2;
       y1 = y_data[n1];
       y2 = y_data[n2];
       vector vd;
       vd = abs( x_data - x3 );
       double xta, xtb;
       vd.GetMinMax( xta, xtb, &n3 );
       y3 = y_data[n3];

    Click Compile button to compile it. Then click Return to Dialog button. Click Finish button.

Define Derived Parameters for Slopes and Intercepts

During function defined process, you can also define some additional derived Parameters such as slopes and intercepts , which are computed from the function parameter values after the fitting process ends.

  1. Click <<Back button twice to go back to Variables and Parameters page, type a1,b1,a2,b2 in the Derived Parameters field.
  2. Click Next button four times to go to Derived Parameters page, fill in the Meaning column and type the equations in the Derived Parameters Equations area as follows, then click Finish button.
    Piecewise slope and intercept.png

Fit the Curve

  1. Select Analysis: Fitting: Nonlinear Curve Fit from Origin menu. In the NLFit dialog, select Settings: Function Selection, in the page select User Defined from the Category drop-down list and pwl2s function from the Function drop-down list.
  2. In the NLFit dialog, select Parameters tab, and fix parameters x1, x2 as shown in the dialog.
    Fit PWL FixP.png
  3. Click Fit button to fit the curve.

Fitting Results

The fitted curve should look like:

Fit PWL FitCurve 2.png

Fitted Parameters are shown as follows.

Parameter Value Standard Error
x1 0.8 0
y1 -0.0271 0.01063
x2 60 0
y2 0.95585 0.0083
x3 22.26316 0.58445
y3 0.66106 0.01197
a1 -0.05275 0.01123
b1 0.03206 8.7153E-4
a2 0.48715 0.01664
b2 0.00781 3.86455E-4

Thus the intersection point for the two segments is (22.26316, 0.66106).

Note that fitting with a piecewise linear function for more than two segments can be done in a similar way.