[R-SIG-Finance] : Comparing Forcasts

[alexios]

For the SPD test of Hansen you'll need to provide a loss series, which 
means you'll need a measure of realized volatility to compare the vol 
forecast to. Alternative approaches would be to use forecast VaR to 
construct a loss series as in Gonzalez-Rivera, Lee, and Mishra (2004) 
who adopt a loss function used in quantile estimation.

You might find the recent model confidence set (mcs) of Hansen, Lunde 
and Nason (2011)  more suitable.

Forecasts from a ugarchforecast object may be extracted through a number 
of methods including as.data.frame, as.array etc  (read the documentation).

Regards,
Alexios

On 20/01/12 19:07, Eric Zivot wrote:
> The easiest test of predictive accuracy is the Diebold-Mariano test (see
> http://faculty.washington.edu/ezivot/econ584/dieboldMariano.pdf and
> http://faculty.washington.edu/ezivot/econ582/econ512forecastevaluationSlides
> .pdf ). This can be implemented via simple regression with a standard error
> correction for serial correlation (e.g. Newey-West std errors). This theory
> for this test is based on known coefficients in your forecasting models and
> non-nested models. There have been many improvements to this test over the
> years. See the many papers by Todd E. Clark and Michael McCracken and Ken
> West. The tests implemented in the ttrTests are in the class of predictive
> accuracy tests.
>
>
> Eric Zivot                  			
> Robert Richards Chaired Professor of Economics and Director of Outreach
> Adjunct Professor of Finance
> Adjunct Professor of Statistics
> Adjunct Professor of Applied Mathematics
> Department of Economics
> Box 353330                  email:  ezivot at u.washington.edu
> University of Washington    phone:  206-543-6715
> Seattle, WA 98195-3330
> www:  http://faculty.washington.edu/ezivot
>
>
>
> -----Original Message-----
> From: r-sig-finance-bounces at r-project.org
> [mailto:r-sig-finance-bounces at r-project.org] On Behalf Of Zachary Mayer
> Sent: Friday, January 20, 2012 5:15 AM
> To: Papa Senyo
> Cc: r-sig-finance at r-project.org
> Subject: Re: [R-SIG-Finance] : Comparing Forcasts
>
> Hi Papa,
>
> I've never used an SPA test, so I don't know if it's implemented in R.
>   Could you point to an article describing it, or an implementation in
> another language?  One idea would be to compare the MAE of your forecast to
> the MAE of a naive forecast (where each value in a time-series is predicted
> to be the same as the previous value).
>
> Whatever you do, it's probably a good idea to make your comparisons
> out-of-sample, and an even better idea to cross-validate.  You can find
> some example code for time series cross-validation here and here:
> http://robjhyndman.com/researchtips/tscvexample/
> http://moderntoolmaking.blogspot.com/2011/12/time-series-cross-validation-3.
> html
>
> -Zach
>
> On Fri, Jan 20, 2012 at 5:21 AM, Papa Senyo<papa.senyo at yahoo.it>  wrote:
>
>>
>>
>>
>> Dear all,
>> Please, I would like to know if the test for superior predictive
>> ability(SPA) is implemented in R
>>
>> Also how can we get the forecast of volatility of return series. Just an
>> information, the SPA compares different forecast and the select the best
>> performance measure. Can this be done using rugarch as it only reports
> MSE,
>> MAE, DAC,
>> I hope to hearing from you all.
>> Kind regards
>> Papa
>>         [[alternative HTML version deleted]]
>>
>> _______________________________________________
>> R-SIG-Finance at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
>> -- Subscriber-posting only. If you want to post, subscribe first.
>> -- Also note that this is not the r-help list where general R questions
>> should go.
>>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-SIG-Finance at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> -- Subscriber-posting only. If you want to post, subscribe first.
> -- Also note that this is not the r-help list where general R questions
> should go.
>
> _______________________________________________
> R-SIG-Finance at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> -- Subscriber-posting only. If you want to post, subscribe first.
> -- Also note that this is not the r-help list where general R questions should go.