On 10/13/2009 4:45:03 AM, Tom_Gutman wrote:
>I don't understand your point. I didn't question this post in the collaboratory, nor the placement in this particular section. I questioned the placement of this post, dealing with an error in a particular implementation of a reduction to echelon for, in an established thread dealing with Cramer's rule. It is a quite different issue, and should start its own thread.
Oh,I'm sorry, I'm bad interpret your previous post. You're right, better, a new thread.
>As to the point of the post, well, what is it? The post merely states that there is an error, borne out by the sheet. Is this just a complaint about the particular implementation (source unknown)?
I assume that this is. About the source, the 'E' procedures are called for 'Elementary row operations'. Taking this as rules, are the better way to solve by hand a linear system (this is, with the augmented matrix) by hand, even more easy -for large systems- than the Cramer rule and gives also the rank of the matrix if eventually the determinant of the system is zero at the same time that reduce the system.
>Or is it a statement that the echelon form solution doesn't work in general?
A further reading of the original post showme that this is a valid aception, isn't clear the proposal of the post, and you're right.
>If indeed intended as commentary of some sort in the particular implemention, what is wanted? The identification of where the error occurs ...
I think so.
> ... (a lot of users seem to have never learned to use the error trace facility)? ...
Please, include-me in this list. I usual forgot this tool because is some uncomfortable to use.
> ... A patch type fix for this particular routine? A general discussion of how to do gaussian elimination in a robust way?
Again, you're right, isn't clear for the collab post.
I post what I found as the answer in
http://collab.mathsoft.com/read?123456,23e#129280but appear in not time order in this thread.
Regards. Alvaro.