Spike analysis case

Question 1

Spikes definition

Answer 1

Spikes are all-or-none events, where information is coded in the timing
of spikes rather than the voltage amplitude

Question 2

Two types of neurons

Answer 2

1. Regular-firing neuron
2. Burst-firing neuron -> neuron repeatedly fires discrete groups or bursts of spikes

Question 3

spike timestamps

Answer 3

timepoint when the neuron spikes (s)

Question 4

spike intervals

Answer 4

• ISIs
• difference between timestamps
• is informative when you want to look at temporal spiking patterns

Question 5

Firing rate

Answer 5

• spikes/s, Hz
• Is informative when you want to look at average firing activity over time, over trials
  (e.g. stimulus repetitions), or over a population of neurons.

Question 6

Aliasing

Answer 6

• is the distortion or artifact that results when a continuous
  signal is reconstructed with finite samples. It suggests structure and information that is present in a signal, while it is not.

Question 7

ways to minimize aliasing

Answer 7

1. using more bins
2. smoothing

Question 8

spike increment

Answer 8

• number of spikes in a discrete time bin.
• The sequence of spike counts across all the bins is referred to as the increment process for a given spike train.

Question 9

Fano factor

Answer 9

• is a measure of the variability in the amount of spikes per time
  bin as a proportion of the mean spikes per time bin.
• We can calculate the Fano Factor as the ratio of the variance
  to the mean spike count

Question 10

FF==1
FF<1
FF>1

Answer 10

• FF == 1: spikes occur like a Poisson process
• FF < 1: spikes are more regular than Poisson (FF is lower for regular spikers)
• FF > 1: spikes is are more variable than Poisson (FF is higher for irregular
  spikers, e.g. bursting neurons)

Question 11

poisson process

Answer 11

mathematical model of spike trains that assumes
that the mean and variance is equal across all time bins.

Question 12

autocorrelation

Answer 12

• cross-correlating a signal with itself with a known lag (L) to find repeated,
  temporal structures.
• In cross-correlation, you shift two vectors and at every shift/lag you compute the
  correlation between one signal and the shifted/lagged version of the other signal.

Question 13

Significance testing of ACs

Answer 13

• Useful to determine whether the spike train is generated from a random process

Question 14

Fitting spike counts.

Answer 14

The number of spikes in a bin follow a Poisson distribution under the Poisson model. Just like the normal distribution, which can be described by two parameters (mean and std), the Poisson distribution can be described with a rate parameter lambda and variable k as the number of spikes in one bin.

Question 15

A KS-plot is simply the observed CDF versus the model CDF:

Answer 15

any deviations outside the 95%
confidence bounds represent a significant deviation between the observed and model distributions.