TABLE 1—
Autoregressive Moving Average (ARMA) Analyses and Estimated Lives Saved: South US Census Region States and All US States and Washington, DC; 1975–2013
| Parameter | South US Census Region States, Estimate (95% CI) | All US States and Washington, DC, Estimate (95% CI) |
| Quartera | ||
| First | 1.70 (1.47, 1.96) | 1.79 (1.58, 2.03) |
| Second | 1.42 (1.23, 1.64) | 1.56 (1.37, 1.78) |
| Third | 1.41 (1.22, 1.63) | 1.48 (1.30, 1.67) |
| Fourth | 1.74 (1.50, 2.03) | 1.71 (1.50, 1.96) |
| Timeb | ||
| Before adoption | 0.010 (−0.453, 0.475) | 0.150 (−0.273, 0.574) |
| After adoption | −0.577 (−1.098, −0.053) | −0.500 (−0.954, −0.043) |
| Comparison groupc | 2.40 (1.56, 3.70) | 4.43 (2.51, 7.82) |
| Autoregressived | ||
| First | 0.195 (0.038, 0.352) | 0.189 (0.032, 0.346) |
| Second | 0.213 (0.057, 0.370) | 0.181 (0.022, 0.340) |
| Third | 0.255 (0.097, 0.413) | 0.247 (0.088, 0.405) |
| Estimated lives savede | 3942 | 3506 |
Note. CI = confidence interval.
Adjusted baseline quarterly fatality rates describing seasonality in Florida pedestrian fatalities, where rates are in number of pedestrian fatalities per 100 000 people at risk. The estimates are adjusted for log fatality rates in the comparison states.
Percentage of change in log fatality rate per quarter before the adoption of Statute 335.065 and percentage of change in log fatality rate per quarter after adoption of the statute relative to the rate of change before adoption. For example, when adjusted for log fatality rates in the South US census region, pedestrian fatality rates in Florida increased by 0.010% per quarter before adoption of Statute 335.065 and decreased by a net total of 0.010%–0.577% = 0.567% more per quarter after adoption of the law.
Risk ratio for the impact of pedestrian fatality rate in comparison states on the pedestrian fatality rate in Florida. The pedestrian fatality rate in Florida is expected to increase 2.4-fold for every 10-fold increase in the South US census region, and the pedestrian fatality rate in Florida is expected to increase 4.43-fold for every 10-fold increase in all US states and Washington, DC. Given that the risk ratio for all US states and DC is higher, it suggests that this comparison group was a better predictor for the temporal pattern of pedestrian fatalities than was the South US census region.
The best-fitting ARMA model for the data required the use of at least 3 past values, which are the autoregressive parameters. The estimates indicate which measurements from preceding quarters are stronger predictors of the current value of each model. As the largest value, the third autoregressive coefficient is the strongest predictor for the current value of the time series for both models. As such, the log fatality rate 3 quarters in the past is the strongest predictor of the current log fatality rate.
Estimated total number of lives saved between adoption of Statute 335.065 in 1984 and 2013.