Readme for burst.exe

20 Sept, 1995

The rationale of our algorithm is presented in the manuscript (Hanes, Thompson and Schall (1995) Relationship of presaccadic activity in frontal eye field and supplementary eye field to saccade initiation in macaque: Poisson spike train analysis. Experimental Brain Research,103:85-96).

This algorithm is an extension of the algorithm utilized in Hanes et al. 1995 that can detect multiple bursts of activity within a spike train. We would be very interested in your comments about the performance of the algorithm on your data.

/***************************************************************************/

To run the program:

at the dos prompt type: burst<enter>



Four inputs will be required...



1. Spike time file. You will be prompted to provide the name of a

text (ASCII) file with spike times sorted in ascending order in a

single column (a <return> between each spike time). The spike time

file can contain up to 1000 spike times. If you need more, tell us.



A sample file called spike.txt is provided as an example.



The spike time file has to be in the same subdirectory as burst.exe.



The spike time file can be named anything you like, within the limits

of DOS.



2. Start time. The time to begin looking for bursts or the beginning of

the spike train.



3. End time. The time to end looking for bursts or the end of the spike

train.



Start time and end time do not have to include the entire range of spike

times in the spike time file. Likewise, start time can be before any spikes

and end time can be after. You can experiment with these times to give the

best results. In effect, by changing these times you are changing the

average spike rate during the time being analyzed. This will change the results.



4. Also, the significance level needs to be specified. All putative 'bursts'

with a surprise index corresponding to a significance level above the level

given by the user will not be listed. For example, p < .05 will limit bursts

to those that are significantly different from a random poisson distribution

of spikes at a .05 level.



Activation times will be listed at the users request. See Hanes et al.,1995

for an explaination of 'activation'.





The C code for the Poisson spike train analysis is saved in burst2.c