Table 5. Pairwise comparison of error matrices between Model 2 and the four simpler models.
Pairwise combination | Z statistic | Significant/Not significant* | |
Model 2 | Model 3 | 9.54 | Significant |
Model 2 | Model 4 | 11.22 | Significant |
Model 2 | Model 5 | 3.91 | Significant |
Model 2 | Model 6 | 0.29 | Not significant |
Model 2 included all variables, while the simpler models contained different combinations of the variables (see Table 3). The significant level indicates whether two error matrices (i.e. from two different models) are completely different (significant), or capable of producing similar results (not significant).
*Significant at the 95% confidence interval (critical value Z = 1.96).