10: Scoring and Statistical Analysis of ERP Amplitudes and Latencies
-
-
Last updated
-
-
Save as PDF
-
Learning Objectives
In this chapter, you will learn to:
-
Score the magnitude of an ERP component using the peak amplitude and the mean amplitude.
-
Score the timing of an ERP component using the peak latency, fractional area latency, and fractional peak latency.
-
Evaluate these different scoring methods in terms of measurement error, effect sizes, and bias.
-
Conduct statistical analyses that minimize the likelihood of bogus-bug-significant results.
-
Correlate ERP measures with behavioral measures.
This chapter focuses on the final steps of data analysis, in which you quantify the amplitudes and/or latencies of your ERPs and conduct a statistical analysis. It doesn’t cover every possible way of scoring amplitudes and latencies, and it barely scratches the surface of the statistical analysis of ERP data. However, it covers the scoring procedures that are used most often (or that should be used most often), along with some very simple statistical analyses. Additional details about scoring can be found in Chapter 9 of Luck (2014), and a much more in-depth treatment of statistical analysis can be found in Chapter 10 of that book. In particular, I encourage you to read about the
jackknife
and
mass univariate
statistical approaches (which are too advanced for the present book).
One reason that I don’t go too deeply into statistical analyses in this chapter is that ERPLAB doesn’t include statistical functions, and I don’t want to have to explain how to use some other statistical package. I’m assuming that you already know how to conduct basic statistical analyses (
t
tests and within-subjects ANOVAs) and have a statistical package that you can use to perform these analyses. If you don’t, I recommend
JASP
(Love et al., 2019), which is free and easy to use. It’s what I used for the analyses in this chapter.
The exercises in this chapter will examine the
lateralized readiness potential
(LRP), which reflects motor preparation. The data are from the ERP CORE flankers experiment. However, the lessons you will learn can be applied to almost any ERP component in almost any paradigm. And the LRP provides excellent opportunities to ask interesting questions about both amplitudes and latencies.
Quantifying amplitudes and latencies is often called the
measurement
process, and in ERPLAB it’s done with the
Measurement Tool
. Recently, however, I’ve started using the term
scoring
instead of
measurement
. When we put electrodes on the scalp and record the EEG, that feels like we’re actually measuring something (the voltages on the scalp). But applying an algorithm to an ERP waveform and hoping that it accurately captures the magnitude or timing of some underlying brain signal doesn’t seem like taking a
measurement
. I now prefer the term
scoring
as more neutral term that is used in many other research areas. For example, you might score the amplitude or latency of a given ERP component.