Rate dependence of unitary event analysis. Roy, A.; Steinmetz, P.; and Niebur, E. Soc. Neurosci. Abstr., 28:1513, November, 1998. abstract bibtex Because the null hypothesis used to compute the chance level of coincidence incorporates a varying firing rate, the occurrence of UEs is independent of the rate of firing when the same level of significance can be applied to each time interval tested. When the rate of firing is low, however, the discrete nature of neuronal firing results in discrete possible significance levels and it becomes impossible to assign similar significance levels to different time intervals. We demonstrate this with data recorded in somatosensory area II. The activity of 30 pairs of cells was recorded during tactile stimulation in a controlled attentional state (Fitzgerald, Lane, Hsiao, this volume) and analyzed using UE analysis. In all pairs, the frequency of UEs changed with the firing rate of the neurons, typically 2-5 spikes/sec. Furthermore, the frequency of UEs in 9 pairs (30%) varied with the attentional state of the animal. The difference between attentional states could be accounted for, however, by the different discrete significance levels. Not only this difference, but indeed the UEs themselves, disappeared when similar significance levels were obtained by using a larger analysis interval (1500 ms). These results suggest that careful interpretation of variations in the occurrence of UEs is required, particularly in the presence of low firing rates. Specifically, when analyzing 150 trials of an experiment using an analysis window 100 ms wide with a 5 ms coincidence window, rates lower than 10 spikes per second require caution.
@article{ Roy_etal98a,
author = {A. Roy and P.N. Steinmetz and E. Niebur},
title = {Rate dependence of unitary event analysis},
journal = {Soc. Neurosci. Abstr.},
year = {1998},
volume = {28},
pages = {1513},
month = {November},
abstract = { Because the null hypothesis used to compute the chance level of coincidence
incorporates a varying firing rate, the occurrence of UEs is independent
of the rate of firing when the same level of significance can be
applied to each time interval tested. When the rate of firing is
low, however, the discrete nature of neuronal firing results in discrete
possible significance levels and it becomes impossible to assign
similar significance levels to different time intervals.
We demonstrate this with data recorded in somatosensory area II. The
activity of 30 pairs of cells was recorded during tactile stimulation
in a controlled attentional state (Fitzgerald, Lane, Hsiao, this
volume) and analyzed using UE analysis. In all pairs, the frequency
of UEs changed with the firing rate of the neurons, typically 2-5
spikes/sec. Furthermore, the frequency of UEs in 9 pairs (30%) varied
with the attentional state of the animal. The difference between
attentional states could be accounted for, however, by the different
discrete significance levels. Not only this difference, but indeed
the UEs themselves, disappeared when similar significance levels
were obtained by using a larger analysis interval (1500 ms).
These results suggest that careful interpretation of variations in
the occurrence of UEs is required, particularly in the presence of
low firing rates. Specifically, when analyzing 150 trials of an experiment
using an analysis window 100 ms wide with a 5 ms coincidence window,
rates lower than 10 spikes per second require caution.},
downloads = {Roy_etal98a.pdf; }
}