I am curious if anyone has had much experience attempting to predict random walks using Loess regression or a variant of local statistical methods.
$\begingroup$
$\endgroup$
2
-
$\begingroup$ It is used in the fitting of volatility surface (local volatility), so you can find some applications there. $\endgroup$ – Magic is in the chain May 30 '20 at 16:41
-
$\begingroup$ a pure random walk cannot be predicted any better than simply taking the last value. So statically any other means would increase both bias and variance. $\endgroup$ – Will Gu Jun 1 '20 at 17:15
Add a comment
|
