Model | Type of travel-volume | GLM | GEE |
---|
Coef1 | p value1 | AUC2 | Coef1 | p value1 | AUC2 |
---|
AT Model 1 | Total | -0.092 | 0.923 | 0.319 | -0.590 | 0.590 | 0.319 |
AT Model 2 | 1st class | -123.541 | 0.244 | 0.457 | -141.454 | 0.360 | 0.457 |
AT Model 3 | 2nd class | -8.622 | 0.629 | 0.329 | -16.912 | 0.550 | 0.329 |
AT Model 4 | Economy class | -0.055 | 0.958 | 0.313 | -0.559 | 0.630 | 0.313 |
- Note: The response variable is whether there is at least one imported case between two specific countries in a specific year. The importation risk index (IR index) was defined as product of travel-volume and disease burden in the source countries. As there were four variables for the travel-volume (number of 1st class passengers, number of 2nd class passengers, number of economy class passenger and total number of all passengers), four IR indices can be calculated. Four logistic regression models and four GEE models were fitted using the four IR indices, respectively as the index was the only independent variable in each model. The IR index variable in these models were rescaled by dividing 10^5 to increase coefficient. P values of Hosmer and Lemeshow goodness of fit in the GLM models were 0.423, 0.583, 0.462, and 0.078 for the model 1–4, respectively, indicating lack of fit for the AT Model 4
- 1coefficients and p values for importation risk index
- 2area under curve for Receiver operating characteristic