Reader Comments

Post a new comment on this article

Amount of data required to compute the newly introduced metrics

Posted by rjolivet on 18 Aug 2009 at 15:35 GMT

Could the authors indicate how many ISIs are required in order to compute LvR or Lv with a fair precision?

No competing interests declared.

RE: Amount of data required to compute the newly introduced metrics

shino replied to rjolivet on 18 Aug 2009 at 23:52 GMT

In our paper, we required the number of spikes > 2000 for the purpose of examining the invariance of the firing irregularity in time, but the instantaneous irregularity can be estimated using fewer interspike-intervals (ISIs).

I may venture to say that the number of ISIs needed to obtain a reliable estimate of firing irregularity is from n=50 to 100: Because the local variation, LvR or Lv, is given as the arithmetic average of independent terms, the metric obeys the central limit theorem. Therefore, its standard deviation (SD) is expected to be of the order of sqrt{1/n}. For n=100, SD is expected to be about 0.1. It should be noted that the conventional coefficient of variation, Cv, is not given in the arithmetic average, and does not obey the central limit theorem, and therefore its SD turns out to be much larger, giving greater fluctuation.

No competing interests declared.