e060824citral_2_lambda

A simple odor response analysis.

We present here briefly the estimation of the conditional intensity of neuron 2 of data set: e060824citral. Before going through this page check the results of the automatic analysis. The actual script used to generate this analysis is attached at the bottom of this web page.

Since the spontaneous activity of the neuron suggests that a renewal process could be adequate for this neuron (in the absence of stimulation), we try out the simplest possible model: an additive model depending on the elapsed time since the last spike (variable TmO) as well as on the time with respect to the stimulus onset. We moreover force the time effect to be uniform prior to the stimulus by setting the time of each bin prior to the stimulus to 6s (the stimulus starts at 6.1s). We write this modified time variable s. The analysis of the distribution of the TmO variable shows that its 5th root is more uniformly distributed. We will then work with the latter that we write: TmO.5th. Our model is therefore symbolically written:

event ~ TmO.5th + s

20 stimulations were applied to the neurons in this data set. In order to validate the model we fit it to stimulations: 4, 8, 12, 16 and 20 and we check it by time transforming the remaining 15 stimulations. The results of the "extended Ogata's tests" are shown next:

Based on these diagnostic plots the model is accepted.

To get a better estimation of the functional form of the two terms of the model we refit the model to the whole data set and get:

We see on the top row the effect of the ellapsed time since the last spike with the refractory period (eraly negative part), a peak at 14 ms and an effect becoming null at 100 ms. The stimulus effect is biphasic excitatory / inhibitory (the white rectangle shows the time during which the valve controling the odor delivery stayed open). The maximal effect of the stimulus is moreover larger (but not much larger) than the maximal effect of the ellapsed time since the last spike.