Figure 3.

Predictor variable importance and interactions as determined by CTree analysis. The CTree algorithm within Package party in R was applied to the ICEES multivariate dataset (see Figure 1 and Table 1) to confirm the most important predictor variables determined by the CRF analysis in terms of relation to the dependent variable of total annual ED or inpatient visits for respiratory issues and to identify significant interactions between predictor variables. The enumeration levels for each variable are indicated on each branch of the tree. Variables are defined in Table 1.