Nonlinear Autoregressive Conditional Duration Models for Traffic Congestion Estimation

Table 3

Estimation results for WACD(1,1) and T-WACD(1,1) models for the observed congestion duration.

Variable	WACD(1,1)	T-WACD(1,1)
		Regime 1 (𝑡^∗≤18 min)	Regime 2 (𝑡^∗>18 min)
Coefficient (st. error)	𝜔0=0.867(0.145)	𝜔0^(1)=0.579(0.149)	𝜔0^(2)=−3.287(1.407)
	𝛾1=0.185(0.035)	𝛾1^(1)=0.542(0.0556)	𝛾1^(2)=0.157(0.055)
	𝜔1=0.542(0.056)	𝜔1^(1)=0.185(0.035)	𝜔1^(2)=0.994(0.332)
	𝑎=0.991(0.026)	𝑎^(1)=0.991(0.030)	𝑎^(2)=1.268(0.325)
AIC *	4.227	3.411
𝑄(10)	10.477	5.156
𝑄(20)	31.610	12.534
Nonlinearity test (4)	1.587(0.082)	1.023(0.503)
LM ARCH test (4)	5.099(0.276)	0.546(0.651)
MAPE (%)	21.68	17.26

* AIC=−2(log𝑙−𝑘)/𝑇, log𝑙is the log-likelihood value, 𝑘 is the number of parameters, and 𝑇 is the number of observations.