Non-Parametric Spectral Tests for Forecast Accuracy

We consider a particular concept of accuracy of predictions, and we develop a class of non-parametric, spectral density tests capable of deciding whether a given random variable can predict a time series. Under standard assumptions, we show that those tests are consistent, robust and admissible. In controlled simulations, we find that those tests rightfully reject with low P-value and high power. When applied to daily returns of NASDAQ Composite (resp. S&P500), we find that an ARMA-GARCH mixture with Normal jumps (resp. T-Student jumps) passes our test at high level of confidence

MoreLess

Year of publication:	2013
Authors:	Leoni, Patrick L.
Publisher:	[2013]: [S.l.] : SSRN
Subject:	Prognoseverfahren \| Forecasting model \| Zeitreihenanalyse \| Time series analysis \| Nichtparametrisches Verfahren \| Nonparametric statistics \| Theorie \| Theory

freely available

Full text |

More access options

doi.org

Check Google Scholar

In German libraries (KVK)

I need help

Which error do you want to report?

Report: URL of this record is broken

Please note: Sometimes more than one URL is available for a resource. Please click on "More Options" to display these URLs

Report: Addition of an URL to that resource

Please note: Sometimes more than one URL is available for a resource. Please click on "More Options" to display these URLs

Report: The language(s) of that resource is/are…

Please note: Required language not in the list? Please use comment field

Report: Addition of an identifier to that resource

Report: report typo

Report: Add statement

Report: Add keywords

Please use the commentary box to describe other errors

Would you like to receive a feedback?

Commentary

Please type the code from the image

For general questions please use this address.

Persistent link: https://www.econbiz.de/10013082748