Approximating Coefficients for Line Approximation from Experimental Data

[Zurtex]

I've got a situation where I can approximate a line by the function:

y = a_0 + a_1 \sin x + a_2 \sin 2x + b_1 \cos x + b_2 \cos 2x

From experimental data I will be able to find certain values for x and y, namely y_n and x_n for some value of n. Now I can easily write a computer program which could work out the co-efficients in this problem if I have 5 values for y and 5 values for x. However, being a particle situation and wanting as many results as possible or perhaps not being able to gain that many results due to constraints, how would I be able to approximate these co-efficient please?

 

[shmoe]

How about least squares approximation?

Let the ith row of an n by 5 matrix M be given by [ 1\ \sin(x_i)\ \sin(2x_i)\ \cos(x_i)\ \cos(2x_i)], and let Y=[y_1,\ldots,y_n]^T.

Then coefficients that will minimize the sum of the squares of the differences of your data points to y = a_0 + a_1 \sin x + a_2 \sin 2x + b_1 \cos x + b_2 \cos 2x will be any column A=[a_0,a_1,a_2,b_1,b_2]^T which satisfies the linear system (M^{T}M)A=M^{T}Y

 

[Zurtex]

Thanks I haven't seen that method yet (well I've researched it but not been able to make sense of it). Thanks a lot, just need to try and make sense of it, but I think I can