# Thread: How to solve this best least squares fit problem

1. ## How to solve this best least squares fit problem

Hello,

I find myself needing to solve a number of best least squares fit problems, and honestly I don't remember the specifics of how it's done. Therefore I'm trying to nail the solution to a single problem before moving on with the rest - and I'd really like some help with that. Here's the description:

It is assumed that there is a coherence between the variables $x$ and $y$ of the form $y=c0+c1*x$. Given the following data (table)
x -1 1 2 4
y 0 1 3 4
we wish to find the best possible values for the coefficients
$c0$ and $c1$.

Task 1: Construct an overdetermined system of equations of the form $A*c=y$ determining the vector $c=(c0,c1)^T$ ( $^T$ meaning transposed).

I immediately imagine the answer to this is simply the following table:
$-1$ $0$
$1$ $1$
$( 2$ $3 ) * ( x$ $y )^T = ( c0$ $c1 )^T$ -- the first part here is a 4x2 matrix, but my formatting is awful - sorry!
$4$ $4$
But that doesn't seem to match the required form. What is the correct answer - and why?

Furthermore, I'm most uncertain as to how the following tasks are solved...

Task 2: Calculate the normal equations $A^T*A*c=A^T*y$.

Task 3: Find $c0$ and $c1$.

Could someone explain how it's done? I could really use some help as I pretty much don't have a clue here.

Lots of thanks!

2. No help to find?

3. ok umm that looks really confusing but I'll take a shot since I'm doing least squares.

you have to find the form ax=b right.

the A form would be like

1 -1
1 1
1 2
1 4

x would be c0 and c1

b would be

0
1
3
4

so part 1 done. Then it wants you to find Atransposed times A, which I assume you know how to do. Then it wants you to calculate y (the least squares solution) for c0 and c1, which you just need to find the inverse of AtA times Aty

4. I actually think I figured it out now... Somewhat:
Calculus.pdf