Hello,
I would very much appreciate it if someone could explain to me the reason why:
Var[ log( 1-f(x) )] =~ [ 1/(1-f(x)) ]^2 Var(f(x))
Thank you!
Delta method.
$\displaystyle Var[g(x)] \approx [ g'(x) ]^2 Var(x)$
Think of it like the Taylor expansion of the variance. What you have above is the second order expansion. Note that you get some nice cancelations in your taylor series.
See: page B-5 of
http://www.phidot.org/software/mark/.../pdf/app_2.pdf
Just got out of my Stats midterm and one of the questions was: Find the approximate variance of your MEM estimator.
I haven't used Delta in a while, and don't think I would have thought of it, if I hadn't been floating it around in my head the night before.
Thanks. You are awesome for asking this question.