/Neuron-Analysis-Exercise

A gentle introduction into the world of neural analysis for economic decision-making tasks

Exercise on how to analyse neurons from an economic decision-making task

A subset of processed data from one of our recent experiments featured in Nature Neuroscience (link). The full dataset (unprocessed) is publicly available here.

The neurons (all recorded from ACC) showcase a range of different response profiles, such as this beautiful positive value coder: Example neuron

Data consists of the following files:

  • x.npy

    • [neurons x trials] matrix of the value of the attended cue
    • Cue is one of 5 different values (0.1, 0.3, 0.5, 0.7, 0.9), sampled with a uniform distribution
  • y.npy

    • [neurons x trials x time] matrix of neural activity during each trial.
      • The first dimension of the matrix will be each neuron (i.e. length 10 as there are 10 neurons)
      • The second dimension will be trials (length 342, the number of trials per session)
      • The third dimension will be each time point (length 250 as it is a 2500 ms window sampled at 10 ms intervals)
    • Cue onset is at 400 ms (i.e. point 40)
    • The firing rate has been smoothed, which is why it's not a binary variable (as raw spike data would be)
  • xy.mat

    • As above but in a Matlab compatible file

Exercise

This is an introduction to the basics of neural data analysis/visualisation. Exploring how individual cells in the primate brain alter their firing rates in response to cues of different value.

  • Load the data
  • For each neuron (i.e. with a for loop)
    • Plot the average firing rate across time for each of the 5 cues (like above)
    • On the same plot, plot the standard error for each cue (don't worry about the fancy shading, or making pretty plots!)
  • Do all neurons have higher firing rates for higher value cues (like in the plot above), or do some do the opposite?

Further questions of interest

If you want, further things you can look for in the data:

  • Quantify each neurons value response by doing a linear regression at each time point
  • Overall, do the population respond positively or negatively to the value of the cue?