Table 1 Most parsimonious (ΔAIC = 0) and competing models [ΔAIC ≤ 2, indicating substantial support (Burnham and Anderson 2002)] for response variables. Where a competing model for a response was the null model, no entry is made in this table. The complete set of results is reported in Appendices 0–0. For each response variable, plot data were analyzed using linear mixed models with “site” as the random factor. Predictor variables tested were the measures of EAB-caused ash decline (ADI, AMI), non-ash stand basal area (BA), subcanopy BA, and shrub cover. Topography (upland or wet) was included as a predictor variable if ADI or AMI was included in the most parsimonious or competing model and there was suitable replication within each topographic category. AICc is the small sample Akaike’s information criterion; ∆AIC is the difference between the AICc of a model and the AICc of the best model, and w is Akaike’s weight. ADI represents ash decline index and AMI represents ash mortality index. Statistics reported are based on likelihood ratio and goodness-of-fit tests

From: Release and suppression: forest layer responses to emerald ash borer (Agrilus planipennis)-caused ash death