Table 8 HLN-corrected Diebold–Mariano...

Table 8

HLN-corrected Diebold–Mariano tests for calibrated Brier loss

Comparison	LLA	LLB	Mean d	DMraw	DMHLN	Sig
IMUGI vs Climatology	0.0291	0.0292	−0.0001	−3.261	−3.260	p < 0.01
IMUGI vs Shumway DTH	0.0291	0.0299	−0.0007	−3.571	−3.569	p < 0.01
IMUGI vs Cox PH	0.0291	0.0298	−0.0007	−3.524	−3.522	p < 0.01

Comparison	LLA	LLB	Mean d	DMraw	DMHLN	Sig
IMUGI vs Climatology	0.0291	0.0292	−0.0001	−3.261	−3.260	p < 0.01
IMUGI vs Shumway DTH	0.0291	0.0299	−0.0007	−3.571	−3.569	p < 0.01
IMUGI vs Cox PH	0.0291	0.0298	−0.0007	−3.524	−3.522	p < 0.01

Note(s): All models are calibrated and evaluated on the full test set. Loss is pointwise Brier score with HAC Newey–West variance estimation and forecast horizon h = 1. The null is H₀: E[loss_IMUGI − loss_benchmark] = 0 against the one-sided alternative that IMUGI has lower calibrated Brier loss. The reported significance category is based on the one-sided HLN-corrected p-value

[ViewLarge]

Sharing Unavailable