Research Article

Intelligent Soft Computing on Forex: Exchange Rates Forecasting with Hybrid Radial Basis Neural Network

Table 5

Unit root tests of USD/CAD.
(a)

Test Original series [ value] 1st differences (returns) [ value]

Augmented Dickey-Fuller(I)−1.017396(I)−30.61353
[0.2781][0.0000]
(II)−1.848666(II)−30.61866
[0.3569][0.0000]
(III)−2.454401(III)−30.60792
[0.3510][0.0000]

Phillips-Perron(I)−1.077154(I)−30.66946
[0.2550][0.0000]
(II)−1.794202(II)−30.68702
[0.3836][0.0000]
(III)−2.415434(III)−30.67642
[0.3712][0.0000]

(b)

TestWindowSpectral estimation method
Bartlett kernelQuadratic spectral kernel
Original seriesReturnsOriginal seriesReturns
(II)(III)(II)(III)(II)(III)(II)(III)

KPSS 
: stationary series
Newey-West2.735123
(0.463)
0.668131
(0.146)
0.074401
(0.463)
0.026211
(0.146)
4.846899
(0.463)
1.158977
(0.146)
0.072329
(0.463)
0.025495
(0.146)
Andrews0.380551
(0.463)
0.150720
(0.146)
0.065257
(0.463)
0.022956
(0.146)
0.306275
(0.463)
0.132197
(0.146)
0.065106
(0.463)
0.022906
(0.146)

Elliot-Rothenberg-Stock 
: unit root
Newey-West30.18086
(3.26)
8.632243
(5.62)
0.358755
(3.26)
0.438046
(5.62)
29.71523
(3.26)
8.977311
(5.62)
0.352926
(3.26)
0.431096
(5.62)
Andrews27.06265
(3.26)
8.375229
(5.62)
0.323259
(3.26)
0.394417
(5.62)
27.29972
(3.26)
8.378941
(5.62)
0.323530
(3.26)
0.394799
(5.62)

(I): model without constant and deterministic trend (5%).
(II): model with constant and without deterministic trend (5%).
(III): model with constant and deterministic trend (5%).