Sunday, February 07, 2010

why be a fox, when it's enough to be a hedgehog?

argh: i hate it when this happens;
some of my students have become paranoid.

while discussing multivariate limits, i distinctly remember telling them .. even writing an algοrithm/pseudocode for them:
  1. can you plug in? if there's no indetermιnate form, then all is well. [1]

  2. try simple trajectories first, like lines of varying slopes. if this gives two directional limits, then nothing more complicated is needed.

  3. otherwise, try something complicated. one of these things will work, but not both:

    • try the squeezε theοrem, but make sure you actually have a correct inequality.

    • try higher-order curves, like parabolas or cubics; the exponents are never larger than the exponents in the problem. [2]
on last week's quiz, there are a host of students that tried horizontal and vertical lines, and subsequently to curves of all sorts .. leading nowhere.

others went, right away, to the squeezε theοrem trick, and writing out false inequalities.
in that class, i even spent time on one example showing a wrong inequality and why it's wrong ..

[sighs]
if they only stuck to the algo ..

[1] i've had very little luck explaining cοntinuity in a calculu∫ class. so when in rome, speak as the gladiatοrial crowds do.

[2] technically, it's not lying if i never give them a problem of that order .. \-:

4 comments:

Leonid said...

I try to be even more concrete.

See the two terms in the denominator?
Set one of them equal to zero - that's a curve/line.
Set the other one equal to zero - another curve.
Set them equal to each other - the third curve.

If all results are the same, switch to proving that the limit exists.

janus said...

give me a few more years of cynicism, L, and i might borrow your algo. q-:

Leonid said...

"See that number to the right and above x? Write it down normal size. Put x next to it. Go back to that number and subtract 1 from it. Put the result to the upper right of x, in a smaller size."

The curve strategy is actually quite reasonable: either one of the terms dominates, or they are comparable - what else is there to look for?

Anonymous said...

I like Leo's method. I may have to try it. I usually have them substitute y=cx^n (n can be real) and then equate exponents with n's. But that still requires math but its relatively successful. (especially when the curve that fails is y=x^(1/3)