Financial Innovation

Table 8 CW-tests on 1-step ahead predictions of GA-NoVaS and P-GA-NoVaS methods

From: A model-free approach to do long-term volatility forecasting and its variants

	P value	GA-NoVaS performance	P-GA-NoVaS performance
2018-AAPL-1step	0.99	0.92	0.89
2019-AAPL-1step	0.08	0.81	0.82
2018-BAC-1step	0.63	0.94	0.93
2019-BAC-1step	0.49	1.05	0.99
2018-TSLA-1step	0.27	0.92	0.86
2019-TSLA-1step	0.22	1.02	0.99
2018-MCD-1step	0.57	0.98	0.94
2019-MCD-1step	0.19	0.96	0.95
2018-MSFT-1step	0.17	0.91	0.96
2019-MSFT-1step	0.47	1.00	0.95
2018-Djones-1step	0.64	0.91	0.91
2019-Djones-1step	0.27	0.96	0.97
2018-NASDAQ-1step	0.51	0.92	0.92
2019-NASDAQ-1step	0.48	0.99	0.94
2018-NYSE-1step	0.31	0.93	0.97
2019-NYSE-1step	0.11	0.91	0.92
2018-S &P500-1step	0.42	0.91	0.88
2019-S &P500-1step	0.32	0.97	0.93
11.2019–10.2020-IBM-1step	0.26	0.80	0.80
11.2019–10.2020-KO-1step	0.01	0.75	0.77
11.2019–10.2020-MCD-1step	0.14	0.51	0.56
11.2019–10.2020-SBUX-1step	0.18	0.67	0.67
11.2019–10.2020-CADJPY-1step	0.07	0.46	0.48
11.2019–10.2020-CNYJPY-1step	0.66	0.78	0.75
11.2019–10.2020-USDCNY-1step	0.36	0.98	0.96
11.2019–10.2020-EURJP-1step	0.19	0.95	0.95
11.2019–10.2020-Djones-1step	0.30	0.56	0.55
11.2019–10.2020-S &P500-1step	0.25	0.59	0.58

The null hypothesis of the CW-test is that parsimonious and larger models have equal MSPE. The alternative is that the larger model has a smaller MSPE. The performances of GA-NoVaS and P-GA-NoVaS methods are calculated as we did in “Real-world data analysis” section, which are relative performances compared with the benchmark method (GARCH-direct method)

Back to article page