mean reversion refresher next, something, current, expect, should

ba736

just a refresher on something i had way wrong.

Quant. AR models

intercept/(1-slope) = mean reverting level.

Anytime the current level is ABOVE the mean reverting level, you should expect the next value to be below the last value.

MiniMe7

It should be noted that if b1 is APPROXIMATELY 1, a unit root exists. I've been burned on practice questions where b1 was 0.99.

llxx

june2009 Wrote:
-------------------------------------------------------
> just a refresher on something i had way wrong.
>
> Quant. AR models
>
> intercept/(1-slope) = mean reverting level.
>
> Anytime the current level is ABOVE the mean
> reverting level, you should expect the next value
> to be below the last value.

If you are asking for confirmation, then that is correct.

NO EXCUSES

kingstongal

a random walk has a slope equal to 1 (bi)
so random walks have no mean eversting level bo/1-1
therefore non convarience stationary
AR model can not be used
need to first difference