Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Quadratic to Linear Fit

  1. Oct 14, 2005 #1
    I have a function [tex]y = \sqrt{ax^2 + bx +c}[/tex], and 2 sets of points [tex]{x_i},{y_i}[/tex] that need to be fit on this curve. First in this problem, I need to somehow convert this nonlinear function into linear and then apply least square methods to determine a,b,c.
    What I came up is ofcouse squaring both sides, removing root. Now I have: [tex] y^2 = ax^2 + bx + c[/tex]. I tried factoring this into [tex]a(x-x_1)(x-x_2)[/tex] but don't think this is better form. I'm not sure how is this even possible, there are 3 numbers to be determined as I can find only 2 equations from least square method (slope & intercept). How to determine a,b,c ? I know how to do it in eg. Mathematica, but I'm writing this as a FORTRAN program, so I need to write exact procedure. I don't know how to do nonlinear fits "by hand".


    P.S. Simpler example of what is supposed to be done (at least I think) is function y = ax^n. Here you just take log of both sides to get: log(y) = nlog(x) + log(a), then calculate log(y_i) and log(x_i) and slope of the line with log(x) and log(y) as variables is n, with intercept log(a).
    Last edited: Oct 14, 2005
  2. jcsd
  3. Oct 16, 2005 #2


    User Avatar
    Science Advisor
    Homework Helper

    Somewhat similar to the Log example, you have 3 var's in the quadratic problem: y^2, x^2 and x. You can calculate a, b and c by inputting these 3 var's into a standard regression software (E.g. SAS); or you can write your own OLS routine to calculate them by using the formula [tex]\hat\beta = (Z'Z)^{-1}Z'(y^2)[/tex] where [itex]\beta = (a,b,c)\text{ and }Z=(x^2,x,1)[/itex].
    Last edited: Oct 16, 2005
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook