Table 2. Variance explained of models for geographic assemblages of mosquitoes.
Predictors |
Southern |
Northern |
Widespread |
|||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Species richness |
Relative abundance |
Species richness |
Relative abundance |
Species richness |
Relative abundance |
|||||||
NY | NJ | NY | NJ | NY | NJ | NY | NJ* | NY | NJ | NY | NJ | |
DDT | 100% | 97% | 32% | 95% | 76%† | 92% | 82% | |||||
DDT use | 18% | 16% | 90% | 13% | ||||||||
URB | 7% | 30% | 18% | 24% | 65% | 82% | 48% | |||||
PCP | 2% | 5% | 4% | >1% | 17% | 100% | ||||||
TMP | 33%† | 10%† | ||||||||||
R2, full model | 0.839 | 0.561 | 0.551 | 0.288 | 0.754 | 0.187 | 0.286 | NA | 0.861 | 0.764 | 0.489 | 0.134 |
DDT, dichlorodiphenyltrichloroethane; NA, not available.
Best regression model goodness-of-fit and variable contributions are shown. The explanatory power of models was measured via a pseudo-R2, calculated as: R2=1−[sum(model residuals)]2/[sum(null model residuals)]2, where the null model has an intercept and autoregressive terms. Significant (P<0.05) variables retained in the final regression model were removed from the model singly and per cent reduction in the resulting pseudo-R2 compared with that of the final model is shown in the table. Higher values indicate variables with higher contribution to the overall goodness-of-fit of the model. The abbreviations designate DDT amount (DDT, Z-scores), DDT use by the mosquito control districts (DDT use), precipitation (PCP, standardized precipitation index), urbanization (URB, human population, in 100,000) and either an average annual November through October (widespread) or January temperature (northern) (TMP, °C). For regression coefficients see Figs 6 and 7.
*No significant predictors.
†Significant main and quadratic terms.