Increase in traffic injury crashes following the 2016 Kumamoto earthquake in Japan: A model comparison

Figures & data

Figure 1. Location of the study area in Japan.

Figure 2. Impact model considered for general linear models.

Table 1. Summary of possible models.

Model	Name	Equations
A	Poisson regression model with Fourier terms	(1), (2), and (4)
B	Negative binomial regression model with Fourier terms	(2), (3), and (4)
C	Poisson regression model with monthly dummy	(1), (2), and (5)
D	Negative binomial regression model with monthly dummy	(2), (3), and (5)
E	SARIMAX	Omitted

Figure 3. Annual changes in injury crashes in Kumamoto city, Japan.

Source: Kumamoto city statistical report (KC 2022).

Figure 4. Annual changes in injury crashes in Fukuoka city, Japan.

Source: Fukuoka city statistical report (FC 2022).

Table 2. Summary of estimated impact periods and injury-crash increase in Kumamoto city.

Model	A	B	C	D	E
Regression	Poisson	Poisson	Negative binomial	Negative binomial	SARIMAX
Seasonal	Fourier terms	Monthly dummy	Fourier terms	Monthly dummy
Sample size	108	108	108	108	95
Modified AIC	1242.6	1021.5	1072.9	1002.3	926.9
Impacted period	2016.4–2020.3	2016.4–2020.2	2016.4–2020.3	2016.4–2020.2	None
RMSE	35.04	22.70	35.13	22.80	23.97
Ljung-Box test p-value	1.94 × 10^−13	0.52	1.26 × 10^−13	0.50	0.02
Crash count change (95% CI)	1,571 (1,215, 1,926)	1,642 (1,164, 2,119)	1,594 (911, 2,277)	1,642 (1,010, 2,273)	None
Change rate (95% CI)	1.19 (1.15, 1.23)	1.20 (1.14, 1.25)	1.19 (1.11, 1.27)	1.20 (1.12, 1.27)	None

Note: AIC: Akaike Information Criteria, CI: confidence intervals.

Figure 5. Modeled and counterfactual injury-crash trends for Kumamoto city.

KC. 2022. Kumamoto City statistical report (Kumamoto-shi-tokeisho); [accessed 2021 Mar 2] https://www.city.kumamoto.jp/hpkiji/pub/List.aspx?c_id=5&class_set_id=2&class_id=2515.

 Google Scholar

FC. 2022. Fukuoka City statistical report (Fukuoka-shi-tokeisyo) [accessed 2022 Jul 1]. https://www.city.fukuoka.lg.jp/soki/tokeichosa/shisei/toukei/toukeisyo/index.html.

 Google Scholar

Code availability

The codes that support the findings of this study are available from the corresponding author upon reasonable request.