2.2. Statistical analysis We used multiple techniques for analysis depending on the data format. For continuous rank data (Objectives 1, 4e5), we used two way analysis of variance (ANOVA) to test for the influence of the two explanatory variables (Objective 1: pest and region; Objective 4: control method and region; Objective 5: attribute and region)
and their interaction. When a model was significant, we used Fisher’s least significant difference (LSD) post hoc test to determine which values were different (Zar, 1999). For nominal responses (Objectives 2e3, 6), we used Fisher’s exact test (i.e., test of independence; Zar, 1999) when we had two nominal variables, and the exact multinomial test (i.e., goodnessof-fit test; McDonald, 2009) when we had one nominal variable. When these tests indicated a significant difference, we used
multiple Fisher’s exact tests or exact binomial tests (McDonald,
2009) to determine which responses were different. We used
a ¼ 0.05 for all tests.