On 02/24/2011 10:10 AM, Stavros Macrakis wrote:
> Depending on the form of your expressions, you may want to try taylor or
> the ratweight scheme to do this sort of thing. Those would both allow
> you to straightforwardly keep (for example) first-order tiny1 and tiny2
> terms, but reduce tiny1^2 or tiny1*tiny2 to 0.
In this case I am only interested in dropping things which are
perturbations to a larger effect. so for example:
stuff*(BIG + small*X + small2*Y) -> stuff*BIG
but
stuff*(small*X + small2*Y) -> stuff*(small*X + small2*Y)
In general, I'm determining whether there exists a big constant effect
by subbing in some estimates of the numerical constants and seeing if
abs(BIG/small) > 100 (for example) hence, in general the need for a
match so that I can grab the pieces and evaluate their numerical size.
I don't think the ratweight and taylor solutions could do this
particular task since whether to drop the perturbations depends on the
existence and numerical size of the constant term.