. 2021 Oct 8;113:47–54. doi: 10.1016/j.ijid.2021.10.007

Table 1.

Summary of the estimated parameters and predictive performances of the proposed models

Model	Parameter	Value	Bias	RMSE	Ranked probability score	Dawid-Sebastiani score
Model 1	$R_{0}$ (Tokyo)	1.74 (1.66–1.82)	0.28	1.65	168.94	479 953
	$R_{0}$ (Osaka)	1.37 (1.32–1.41)
	$R_{0}$ (Aichi)	1.15 (1.11–1.19)
	$R_{0}$ (Hokkaido)	1.33 (1.25–1.41)
	Google mobility	0.02 (0.01–0.02)
Model 2	$R_{0}$ (Tokyo)	4.39 (4.05–4.72)	-1.06	2.29	264.41	328 849
	$R_{0}$ (Osaka)	3.23 (3.02–3.44)
	$R_{0}$ (Aichi)	2.59 (2.43–2.75)
	$R_{0}$ (Hokkaido)	2.09 (1.95–2.23)
	Google mobility	0.03 (0.03–0.03)
	Temperature	−0.03 (−0.03 to −0.03)
Model 3	$R_{0}$ (Tokyo)	4.40 (4.07–4.73)	0.55	1.77	171.60	147 027
	$R_{0}$ (Osaka)	2.85 (2.62–3.08)
	$R_{0}$ (Aichi)	2.20 (2.02–2.38)
	$R_{0}$ (Hokkaido)	1.99 (1.84–2.15)
	Google mobility	0.03 (0.03–0.03)
	Temperature	−0.02 (−0.02 to −0.01)
	Risk awareness	−0.12 (−0.15 to −0.10)

Model 1: Google mobility; Model 2: Google mobility and temperature; Model 3: Google mobility, temperature, and risk awareness; RMSE: root-mean-square error

Each value describes the estimated parameters of three different models and their 95% confidence intervals derived by maximum likelihood estimation and Laplace approximation normal distribution. Predictive performances of each model were compared using three measures (root-mean-square error, ranked probability score, and Dawid-Sebastiani score) and by summing estimates of four regions in Japan (prefectures for the training data — Osaka, Tokyo, Aichi, and Hokkaido).