Poisson noise

[ttpuff]

I try to add Poisson noise with some firing rate. I've found people have discussed this issue before here. Basically, NetStim is good choice to generate Poisson spike trains. but I have two questions:

1.) The only problem for NetStim is that you have to assign the time for the first spike. In my case, I want to add Poisson trains for each neuron of the network. So I have to assign the first spike time for N neurons. This will be a problem for me since my model is very sensitive to any spikes' timing. Is there any way to generate Poisson trains from the very beginning of the first spike?

2.) I found my statistics of ISIs looks not good. There are some discrete peaks around every 100ms. When the rate is high, say 10hz, there are four huge peaks at 100, 200, 300, 400ms. what could be the reason for this bug?

thanks a lot.

[ted]

1.) The only problem for NetStim is that you have to assign the time for the first spike.

No. The parameter .start is not the time of the first spike. It is the time at which the probability density function (PDF) is maximum. Before that time, the PDF is 0, and after that time, it decays according to a 1st order exponential.

2.) I found my statistics of ISIs looks not good. There are some discrete peaks around every 100ms. When the rate is high, say 10hz, there are four huge peaks at 100, 200, 300, 400ms. what could be the reason for this bug?

What ISI do you mean? The ISI of the events generated by the NetStim, or the ISI of the spikes generated by the cells in your model? And where is the evidence that there is a bug? Have you considered the possibility that the cells in your model may have a preferential firing frequency, or (if there is a network) that the duration of excitatory or inhibitory events may favor a particular firing frequency?

[ttpuff]

No. The parameter .start is not the time of the first spike. It is the time at which the probability density function (PDF) is maximum. Before that time, the PDF is 0, and after that time, it decays according to a 1st order exponential.

Right, I just notice that. Actually I am using the old-fashion model like the one by Z. Mainen and A. Destexhe. the parameter .start there is the actual time of the 1st spike.

What ISI do you mean? The ISI of the events generated by the NetStim, or the ISI of the spikes generated by the cells in your model? And where is the evidence that there is a bug? Have you considered the possibility that the cells in your model may have a preferential firing frequency, or (if there is a network) that the duration of excitatory or inhibitory events may favor a particular firing frequency?

Since I am not using NetStim, it's the ISI of the spikes generated by the cells in the network. When I say the bug, just because that 100ms peaks look not good. Should it be more continuous with small intervals? I do the same thing in Matlab, that looks OK.

[ted]

I'm confused.

I am using the old-fashion model

The old-fashioned model of what? Used by Mainen and Destexhe where?

it's the ISI of the spikes generated by the cells in the network.
. . .
I do the same thing in Matlab, that looks OK.

If you're saying that your Matlab implementation of your model produces results that differ from those generated by your NEURON implementation, then there is a difference between the two implementations. To discover what is different, you need to verify that all biophysical mechanisms, anatomical specifications, distributions of ion channels etc., and synaptic connections in your NEURON implementation are identical to those in your Matlab implementation.