testing forecasts of VAR and BVAR models

[sanjeev]

is it possible to use the Modified DM test for comparing the forecast performance of a VAR and BVAR model where the macroeconomic model is the same. That is the same model with the same variables is estimated via VAR and then via BVAR.

[TomDoan]

Yes. The forecast errors don't collapse in population like they do in the nested model case.

[sanjeev]

Is it possible to compare the VAR and BVAR model when the variables in the models are the same but the lags differ. That is can we use Modified DM for comparing a VAR of say order 2 with a BVAR of order 3 or is this a case of nested models.

[TomDoan]

Yes. The "B" part still eliminates the theoretical problem generated by the nesting.

[sanjeev]

Dear Tom,
Thank you so much for your help. I have another question. I would be grateful if you could possibly guide me again.
I am forecasting the futures prices of wheat. I employing alternative forecasting models. I am finding out point and interval forecasts, and also interesting in finding out trading strategy based on the point forecasts. I need to construct combination forecasts, testing for economic significance: measures of performances. Can I be able to do the above estimations in rats 9.0.
Actually I am replicating the Eirini and Skiadopoulos 2011 paper titled “Are VIX futures prices predictable? An empirical investigation”. Could you please help me with the code of this paper.
Thanks, Sanjeev

[TomDoan]

You should be able to do that with RATS---there's nothing particularly complicated about the calculations, but simulated real-time trading takes a lot of organization. How far have you gotten?

The paper's behind a paywall, but the abstract seems to indicate that they beat this up and found that no trading strategy worked.

[sanjeev]

You should be able to do that with RATS---there's nothing particularly complicated about the calculations, but simulated real-time trading takes a lot of organization. How far have you gotten?

The paper's behind a paywall, but the abstract seems to indicate that they beat this up and found that no trading strategy worked.

Thank you so much Tom. I am having access to the Eirini and Skiadopoulos 2011 paper. It would be kind if you provide me the code of this paper or related paper. Regards Sanjeev

[TomDoan]

We don't have any. Most papers like this use proprietary data.

[sanjeev]

Thank you so much Tom. I have another question. It would be very kind of you if you could possibly help me with the code for rolling Granger Causality. I want to plot the graph for rolling Granger causality by keeping F-statistics on Y-axix and time on X-axix. Thanks Sanjeev .

[TomDoan]

Almost anything "rolling" is done using the process described in the User's Guide. Inside the loop, you do your causality test, controlling the sample range, save the test statistic, and graph it outside. The ONEBREAK.RPF example which is worked through in considerable detail in that section, does exactly that type of analysis, just a stability test rather than a causality test.

[sanjeev]

Dear Tom,
Thank you so much for your help. I have another question. I would be grateful if you could possibly guide me again.
I am forecasting the futures price of wheat by arma (1,1), arma (1,2) and economic model. The profitability of the trading strategies in commodity futures is evaluated in terms of the Sharpe Ratio and Leland’s (1999) alpha. I also like to evaluate the statistical efficiency of the bootstrapped interval forecasts. I want to report the percentage of observations that fall outside the bootstrapped intervals, and the values of Christoffersen’s (1998) likelihood ratio test of unconditional coverage .
Could you please help me with the code for Sharpe Ratio, Leland’s alpha and Christoffersen’s likelihood ratio test.
Thanks
Sanjeev

[TomDoan]

Aren't the Sharpe ratio and the alpha computed as functions of parameters in simple regressions?

See the GARCHBACKTEST.RPF example for an analysis of trading strategies.

[sanjeev]

Dear Tom,
Thank you so much for your help. I have another question. I would be grateful if you could possibly guide me again. I would like to test the efficiency of interval forecast by Christoffersen’s (1998) test statistic. Could you please help me with the code.
Thanks,
Sanjeev

[TomDoan]

How far have you gotten? It would also be greatly appreciated if you would provide complete references.

[sanjeev]

Dear Tom, Could you please help me with the code of Clark West test (2007). Clark west test is used to compare the forecast from two nested models.
Thanks,
Sanjeev