Psychology 321: Cognitve Psychophysiology Lab.

1

• Psychology 321 Cognitve Psychophysiology Lab.
• jp-rosenfeld_at_northwestern.edu
• annecward_at_gmail.com

2
A simple neural code
3
Event-related potentials
4
Spontaneous EEG
5
Set-Up for ERPs
6
The P300 ERP to meaningful vs. non-meaningful
itemshow to measure?
7
P300 peaks and windows
8
How do you tell if Blue gt Green ?
9
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,
  heres how we make them

10
(No Transcript)
11
For example.

• C\Users\rosenfeld\Desktop\erpav.gif

12
Recall How do you tell if Blue gt Green ? T-test
too noisy for individuals, must use bootstrap
13
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.
• Lets average all these and say we get 1.4.
  Thats the real or actual sample mean of all 30
  balls, each and every one.

14
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?

15
Let us suppose we repeat these 30 set
re-samplings (iterations) 100 times, so thatwe
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,

16
In ERP Bootstrapping..

• ..the original set of single sweeps is
  repeatedly randomly sampled but with
  replacement
• yielding multiple average ERPs in a single
  subject.
• Lets say there are 6 repetitions of sampling of
  18 single sweeps

17
Each set of 18 single sweeps is averaged yielding
6 averages
18
You do this with the blue and green (from above)
erps, except you look at blue green P300 values
(Mx or Dx), and you insist that 90 or more
differences must begt0 to say,yes blue is gt
green.