Table 1.
Overall performance scores of models.
| Measures | Unified model | Variable signal | Noise |
|---|---|---|---|
| SPIKES | |||
| Variable rate | Rate | ||
| AIC | 5 | 0 | 0 |
| CDF plots | 5 | 0 | 0 |
| Error rates on single trials | 5 | 3 | 3 |
| False alarm rate | 2 | 2 | 3 |
| FIELDS | |||
| VSPN | SPN | ||
| AIC | 5 | 0 | 0 |
| K–S test on residual (p > 0.05) | 2 | 0 | 1 |
| Error rates on single trial | 4 | 4 | 5 |
| False alarm rate | 1 | 3 | 1 |
| Total | 29 | 12 | 13 |
Spike-field models are scored according to various reliability measures in five sessions. A score of five means a model is acceptable in five sessions according to the corresponding reliability measure, four means in four sessions and so on. For AIC, False alarm rate and CDF plot, the model with lowest quantitative value takes all points. If there is a tie, a point is given to both models. An acceptable model is considered for K–S tests when p values > 0.05 and error rates for correct visual response onset/movement direction selection are less than chance level.