العربية

Analysis

From clean epochs to results: ERP waveforms, the power spectrum and time-frequency analysis.

5 lessons: 2 core, 2 recommended, 1 optional

After this station you can:

  • Compute an ERP for a condition and read its components
  • Compute a PSD and interpret the alpha peak
  • Measure ERP amplitude and latency reproducibly

Lessons

  1. ERPs & averagingCore15 min
  2. Power spectrum (PSD)Core12 min
  3. ERPLAB: measuring amplitude & latencyRecommended15 minComing soon

    In this lesson you will: Measure the mean amplitude in a predefined time window for many ERPsets at once, from the GUI and from a script.

    Read about it now in the official docs (opens in a new tab)

    This lesson is being written.

  4. Time-frequency: ERSP & ITCRecommended15 min
  5. Data quality (SME)Optional12 minComing soon

    In this lesson you will: Show the aSME values of an ERPset in the ERPLAB table, and read them for each channel and time window.

    Read about it now in the official docs (opens in a new tab)

    This lesson is being written.

Test yourself

4 questions. Your score stays with you and isn't sent anywhere.

1. Why does the ERP show up in the average but hardly in a single trial?
Show the answer

The correct answer: Because the response repeats with the same shape and timing in every trial, while other activity is random and cancels out in the average

A point-by-point average keeps what repeats with the same timing and weakens what varies at random from trial to trial. That is why the waveform gets clearer with more trials.

Review the lesson: ERPs & averaging

2. You raise the number of trials in the average from 20 to 80, four times as many. Roughly what happens to the noise in the average?
Show the answer

The correct answer: It becomes half

Noise in the average falls with the square root of the number of trials, and the square root of 4 is 2, so it halves. That is why a small gain needs many trials, and why similar trial counts per condition matter.

Review the lesson: ERPs & averaging

3. Which statements about reading a spectrum in EEGLAB are true? (Select all that apply)
Show the answer

The correct answer: A gradual fall in power as frequency rises is normal, A 10 dB difference means ten times more power, A narrow peak at 50 or 60 Hz in uncleaned data is a sign of line noise

The vertical axis is logarithmic, so a difference in dB is a ratio, not units. The default range stops at 25 Hz, so 50 and 60 are not visible; widen it. After a 30 Hz low-pass everything above it is gone, so compute the spectrum before it.

Review the lesson: Power spectrum (PSD)

4. In an ERP plot from another program, the largest peak points down. Before you say it is negative, what do you check?
Show the answer

The correct answer: The direction of the vertical axis: is positive up or negative up

Some books and programs plot negative up. In EEGLAB menus positive is up (the 'ydir', 1 option), but check the axis label of every plot before you say "positive peak" or "negative".

Review the lesson: ERPs & averaging