Table 3.
Pairwise t-test results for the convergences derived from the experiments shown in Figure 8.
| β1 = 0.1 | β2 = 0.3 | β3 = 0.5 | β4 = 0.8 | β5 = 1.0 | |
|---|---|---|---|---|---|
| β1 = 0.1 | 0.00 | 0.69 | 0.85 | 0.96 | 1.00 |
| β2 = 0.3 | 0.69 | 0.00 | 0.54 | 0.92 | 0.96 |
| β3 = 0.5 | 0.85 | 0.54 | 0.00 | 0.65 | 0.96 |
| β4 = 0.8 | 0.96 | 0.92 | 0.65 | 0.00 | 0.46 |
| β5 = 1.0 | 1.00 | 0.96 | 0.96 | 0.46 | 0.00 |
For each pair of convergences matching by number of behaviors J ∈ [2, 29] between experiments with different values β, a t-test was performed. This table shows the fraction of points for which the improvement was statistically significant (significance level of 0.05). The analysis shows that the improvement becomes more significant the higher the difference Δβ = βj − βi is.