Table 2.

Desirable properties for a burst detector

D1	Deterministic: The method should detect the same bursts over repeated runs on the same data, to ensure consistency and reproducibility of results.
D2	No assumption of spike train distribution: The method should not assume that ISIs follow a standard statistical distribution, to ensure wide applicability to a variety of spike trains.
D3	Number of parameters: The method should have few parameters, to reduce the variability inherently introduced through parameter choice.
D4	Computational time: The method should run in a reasonable amount of time on standard personal computers.
D5	Nonbursting trains: The method should detect few spikes as being within bursts in spike trains containing no obvious bursting behavior.
D6	Nonstationary trains: The method should detect few spikes as being within bursts in spike trains with nonstationary firing rates that contain no obvious bursting behavior.
D7	Regular short bursts: The method should detect a high proportion of spikes in bursts in spike trains containing short, well-separated bursts.
D8	Nonstationary bursts: The method should detect a high proportion of spikes in bursts in spike trains containing bursts with variable durations and numbers of spikes per burst.
D9	Regular long bursts: The method should detect a high proportion of spikes in bursts and an accurate number of bursts in spike trains containing long bursts with low within-burst firing rates.
D10	High frequency bursts: The method should detect a high proportion of spikes in bursts and an accurate number of bursts in spike trains containing a large number of short bursts.
D11	Noisy train: The method should classify a high number of within-burst spikes as bursting and a low number of interburst spikes as bursting in spike trains containing both bursts and noise spikes.