Psychology 321 Cognitve Psychophysiology Lab jprosenfeldnorthwestern edu annecwardgmail

§Psychology 321: Cognitve Psychophysiology Lab. §jp-rosenfeld@northwestern. edu §annecward@gmail. com

A simple neural code

Event-related potentials

Spontaneous EEG…

Set-Up for ERPs

The P 300 ERP to meaningful vs. non-meaningful items—how to measure?

P 300 peaks and windows

How do you tell if Blue > Green ?

Usually a t-test will work if these averages represent groups of 10 -20 persons… But in diagnostic psychophysiology, we want to compare 2 individual averages, , , by the way, here’s how we make them:

For example…. §C: UsersrosenfeldDesktoperpav. gif

Recall: How do you tell if Blue > Green ? T-test too noisy for individuals, must use bootstrap…

Suppose you have a big bowl of say 30 hollow balls, each with a number varying say 1 to 30. …. § …and inside each is a piece of paper with a number written on it, a number varying from say -50 to +50: § -50, -49, -47……. +46, +48, +49. § Let’s average all these and say we get 1. 4. That’s the real or actual sample mean of all 30 balls, each and every one.

Now we begin the bootstrap process of sampling with replacement: § We stir up the balls, and draw one. We note the number inside and add it to a growing sum that now has one value. § WE REPLACE THE BALL, AND STIR THEM UP AGAIN, AND RANDOMLY DRAW ONE. We note the number inside and add it to a growing sum that now has one sum of 2 numbers. § We repeat this until we have drawn 30 balls with replacement. We divide the sum of 30 by 30 to get the average. How likely is it that it will be 1. 4, the real sample mean?

Let us suppose we repeat these 30 set re-samplings (iterations) 100 times, so that we can now compute an average of these 100 averages…. §Are we gonna get closer to 1. 4? §For sure; according to Efron (1979), as the iterations go to infinity, their average approaches the sample mean,

In ERP Bootstrapping…. . § …. . the original set of single sweeps is repeatedly randomly sampled – but with replacement— …yielding multiple average ERPs in a single subject. § Let’s say there are 6 repetitions of sampling of 18 single sweeps:

Each set of 18 single sweeps is averaged yielding 6 averages…

You do this with the blue and green (from above) erps, except you look at blue – green P 300 values (Mx or Dx), and you insist that 90% or more differences must be>0 to say, ”yes’” blue is > green.