Page 1 of 1

rolling walk-forward optimization for an AR model

Posted: Wed May 14, 2014 5:15 am
by ac_1
Guys,

Is there a way to do a rolling walk-forward optimization as depicted in:

http://www.tradestation.com/trading-tec ... -optimizer

in RATS e.g for an AR model?


Regards,
Amarjit

Re: rolling walk-forward optimization for an AR model

Posted: Wed May 14, 2014 5:36 am
by TomDoan
That's an ad, not an algorithm. However, it sounds like it's just using a holdback period, or simulated out-of-sample forecasts, both of which are routinely done in macroeconometrics. It's what the @RUNTHEIL procedure in chapter 7 of the User's Guide is doing. The CANMODEL.RPF, ARIMA.RPF and SIMULTHE.RPF examples all do something similar.