Comparison of models predicting <i>Varroa</i> intensity, without spatial component. For each variable, we fit an additive model and interaction with time; the model preferred by AIC is presented here. This table includes the results of the additive model for Floral quality, Insecticide, and Nesting quality. The remaining models include the indicated variables and interaction with time. Statistics presented are: Akaike Information Criterion (AIC), % of Deviance explained (a metric similar to but more appropriate for negative binomial regression), Relative Error (RelErr = Mean(), Mean Absolute Error (MAE = Mean()). Relative error is computed relative to the baseline non-spatial model (time only). The best model for each statistic is bolded. There is no clear winner among the models for all criteria, however all the highlighted models include Management as a variable.

Comparison of models predicting Varroa intensity, without spatial component. For each variable, we fit an additive model and interaction with time; the model preferred by AIC is presented here. This table includes the results of the additive model for Floral quality, Insecticide, and Nesting quality. The remaining models include the indicated variables and interaction with time. Statistics presented are: Akaike Information Criterion (AIC), % of Deviance explained (a metric similar to but more appropriate for negative binomial regression), Relative Error (RelErr = Mean(), Mean Absolute Error (MAE = Mean()). Relative error is computed relative to the baseline non-spatial model (time only). The best model for each statistic is bolded. There is no clear winner among the models for all criteria, however all the highlighted models include Management as a variable.