How to do spectral estimation of a point process

I would like to analyse neural firing in the time-frequency domain. I have found that several programs (Chronux, nSTAT) exist for this problem, but they are used for analysis of neural data over multiple trials. I would like to analyse data in which I have only one recording from one particular position. I have found that the spectrum of point process can be expressed as the Fourier transform of its autocovariance function. Can somebody tell me which command should be used for it in Matlab?

Answer 1

[Star Strider]

For time-frequency analyses, I usually use the pspectrum function with the 'spectrogram' option.

[Bence Laczó]

I also use it sometimes, but I don't really know how it can be used for point process data.

[Star Strider]

I have no idea what your data are.

If you have matrix data, perhaps simply using surf would work to visualise it.

[Bence Laczó]

My data are spike trains. It contain zeros and ones (where there is a spike) or I also have the interspike interval values in milisecs.

[Star Strider]

If I remember correctly, that is a Poisson process, so perhaps something like this —

ISI = poissrnd(rand(1,250)*10)*1E-3;
Fs = 1E3;

figure
pspectrum(ISI, Fs, 'spectrogram')
colormap(turbo)

I have never analysed spike trains, so I don’t know the accepted procedures for analysing them.

.

Answer 2

[Image Analyst]

So you have a signal emanating from a point? So you have time series data. I'd just use fft or pwelch. It doesn't matter if your signal came from only one location.

[Bence Laczó]

I have interspike interval data or spike locations which are marked by ones in an array of zeros.

Now I see what was misleading. I have neuronal firing data from different positions in the subthalamic nucleus. There was no specific task during the recording only resting state data.