Table 3. Optimal generalized additive models built with time lags and non-time lags.
| Model type | Formula | GCV | R 2 |
|---|---|---|---|
| No time lag model | |||
| Dependent variables | |||
| Reported incidence | f0+s(SDt) + RHt + s(ATt) + PRCPt + s(Montht) + WSt + s(TDt) | 4.05E-15 | 86% |
| Coefficient of infection of human to human (β1) | f0+SDt + RHt + s(ATt) + PRCPt + s(Montht) + s(WSt) + s(TDt) | 1.59E-19 | 62% |
| Coefficient of infection of environment to human (βw1) | f0+SDt + s(RHt) + s(ATt) + s(PRCPt) + s(Montht) + s(WSt) + s(TDt) | 9.04E-18 | 86% |
| Coefficient of infection of tick (β21) | f0+SDt + RHt + s(ATt) + PRCPt + s(Montht) + s(WSt) + s(TDt) | 5.92E-22 | 96% |
| Coefficient of infection of animal to human (β31) | f0+SDt + s(RHt) + s(ATt) + s(PRCPt) + s(Montht) + s(WSt) + s(TDt) | 3.33E-18 | 75% |
| Time lag 30 days model | |||
| Dependent variables | |||
| Reported incidence | f0+SDΔt + RHΔt + s(ATΔt) + PRCPΔt + s(MonthΔt) + WSΔt + TDΔt | 4.87E-15 | 83% |
| Coefficient of infection of human to human (β1) | f0+s(SDΔt) + s(RHΔt) + s(ATΔt) + s(PRCPΔt) + s(MonthΔt) + s(WSΔt) + TDΔt | 1.76E-19 | 58% |
| Coefficient of infection of environment to human (βw1) | f0+s(SDΔt) + s(RHΔt) + s(ATΔt) + PRCPΔt + s(MonthΔt) + s(WSΔt) + s(TDΔt) | 8.36E-18 | 87% |
| Coefficient of infection of tick (β21) | f0+s(SDΔt) + s(RHΔt) + s(ATΔt) + PRCPΔt + s(MonthΔt) + s(WSΔt) + TDΔt | 6.05E-22 | 96% |
| Coefficient of infection of animal to human (β31) | f0+s(SDΔt) + s(RHΔt) + s(ATΔt) + PRCPΔt + s(MonthΔt) + s(WSΔt) + TDΔt | 3.69E-18 | 72% |
*SD = Sunshine duration, RH = Relative humidity, AT = Average temperature, PRCP = 24-hour precipitation, WS = Wind speed, TD = Tick density, and GCV = generalized cross-validation score