using LA to solve linear least squares

[Vegan]

I was looking at some old FORTRAN and I was disgusted by the ugly code
posted around. So I am fairly good with linear algebra so this is my
idea.

Le A be a matrix and w and b be vectors

So Aw = b is a vector equation to solve the least squares

Let r be a vector

Then r = b - Aw

Gives the residual vector

I am populating A with the col vector of 1, the col vector of x and
the col vector of x^2 and b is the col vector of the y values

Anyone able to prove this?

[Zoltán Tóth]

Man, what are you doing? Earlier i thought that the problem may be with your English knowledge, but now i see, that you are posting messages which are full of logical errors in the explanation. How do you expect a community to help you with negligence in all your sentences. Is this a psychological experiment of yours about how to drive mathematicians crazy?

My point:
1) In the last sentence you ask for mathematical proof. Then why do you mention fortran in you first sentence??? What is a programming language to do with a mathematical proof???
2) How a property of a source code can implicate that you are good in LA?
3) "Let r be a vector. Then r = b - Aw" I do not see an implication here. Why don't you just write "Let r = b-Aw"?
4) Why do you introduce a name ("r") if you do not use it anywhere later?
5) No, nobody can prove that you are populating matrix A with those values.

OK, i can __guess__ that you want to fit a quadratic implicit polynomial onto a set of points. I also have faced a similar problem, and the dissertation of Bo Zheng "2d Curve and 3d Surface representation with using Implicit Polynomial and its Applications" helped me a lot.

Please do not post here until you learn to write.

--
You received this message because you are subscribed to the Google Groups "fcla-discuss" group.
To post to this group, send email to fcla-d...@googlegroups.com.
To unsubscribe from this group, send email to fcla-discuss...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/fcla-discuss?hl=en.

[Vegan]

Thanks for that reference, but I did not find what I was hoping for.

My goal is least squares linear regression using a vector approach to
compute the coefficients as well as the residuals for error
assessment.

Seems my post was truncated slightly, sorry about that.