Statistical tests
| Test use | Test | Data structure | Power |
|---|---|---|---|
| Individual neuron color preferences (Fig. 2) | ROC area significance bootstrapping | Non-Gaussian | 95% CI > 0.5 |
| Population percent color preference (Figs. 2B, D, 3D) | χ2-proportion test | Proportion | p < 0.05 |
| Individual neuron cue/foil preferences | ROC area significance bootstrapping | Non-Gaussian | 95% CI > 0.5 |
| Unit type category AMI comparison (Fig. 3B) | ANOVA | Gaussian | α = 0.05, p < 0.05 |
| Population percent cue/foil preference (Fig. 3C) | χ2-proportion tests | Proportion | p < 0.05 |
| Time course of cue and color preferences (Fig. 3C,D) | ANCOVA | Gaussian | α = 0.05, p < 0.01 |
| Presence of significant change-detection activity in single neurons | Wilcoxon rank sum test | Non-Gaussian | p < 0.05 |
| Comparison of onset-evoked and change-evoked peak firing rates | ANVOA | Gaussian | α = 0.05, p < 0.05 |
| Individual neuron saturation increment vs. decrement preferences (Fig. 5) | ROC area significance bootstrapping | Non-Gaussian | 95% CI > 0.5 |
| Individual neuron hits vs. misses preferences (Fig. 6) | ROC area significance bootstrapping | Non-Gaussian | 95% CI > 0.5 |
| Unit type category detect-probability comparison (Fig. 6) | ANOVA | Gaussian | α = 0.05, p = 0.77 |
| Unit type category percent hits preference comparison (Fig. 6) | χ2-proportion tests | Proportion | p > 0.05 |
| Comparison of hinge model component contributions to variation in threshold time | ANOVA | Gaussian | α = 0.05, p < 0.01 |