Skip to contents

Event-Related SCR Analysis

Source: vignettes/eda-event-related.Rmd
eda-event-related.Rmd

Introduction

Event-related skin conductance response (ER-SCR) analysis examines phasic EDA changes that are time-locked to experimental events (e.g., stimulus presentations, decisions, or feedback). This is the most common paradigm in psychophysiology research using EDA.

This vignette demonstrates how to use PhysioEDA to perform event-related SCR analysis, from data preparation through statistical summary of event-locked responses.

Overview of the ER-SCR Workflow

The typical event-related SCR analysis pipeline consists of:

  1. Load or simulate EDA data with event markers
  2. Preprocess the signal (filter, artifact correction)
  3. Decompose into tonic and phasic components
  4. Extract event-related SCR features
  5. Summarize and analyze results

Preparing Data with Events

For event-related analysis, the PhysioExperiment object must contain event markers. These can be imported from experimental software or added manually.

library(PhysioEDA)

# Generate simulated EDA with 4 stimulus events
x <- make_eda_with_scr(n_time = 6000, sr = 10, n_events = 4)

# Inspect the events
events <- getEvents(x)
events

Adding Events Manually

If your data does not already contain events, you can add them using setEvents():

library(PhysioCore)

x_plain <- make_eda(n_time = 6000, sr = 10)

# Define stimulus events
x_plain <- setEvents(x_plain, PhysioEvents(
  onset = c(60, 120, 240, 360),
  duration = c(0.5, 0.5, 0.5, 0.5),
  type = c("stimulus", "stimulus", "stimulus", "stimulus"),
  value = c("CS+", "CS-", "CS+", "CS-")
))

Preprocessing for ER-SCR

The preprocessing steps are the same as for general EDA analysis: filtering and artifact correction. It is important to complete preprocessing before decomposition.

# Lowpass filter to remove noise
x <- edaFilter(x, type = "lowpass", cutoff = 1.0)

# Detect and correct artifacts
x <- edaArtifact(x, correct = "interpolate")

Decomposition

For event-related analysis, the Continuous Decomposition Analysis (CDA) method is recommended as it provides a physiologically meaningful driver signal.

x <- edaDecompose(x, method = "cda", tau1 = 0.75, tau2 = 2.0)

# Verify that tonic and phasic assays exist
SummarizedExperiment::assayNames(x)

The edaErscr() function is the core of event-related analysis. For each event, it searches for an SCR onset within a specified latency window and extracts amplitude, latency, rise time, and 50% recovery time.

Understanding the Parameters

  • onset_window: The valid latency range (in seconds) after event onset in which an SCR onset must occur to be considered event-related. The default c(1, 4) reflects the typical 1-4 second latency of sympathetic responses.
  • peak_window: The minimum and maximum time from SCR onset to its peak (default: c(0.5, 5)).
  • amplitude_min: The minimum amplitude threshold for a valid SCR (default: 0.01 microsiemens).
# Extract event-related SCRs
er_results <- edaErscr(x,
  event_type = "stimulus",
  onset_window = c(1, 4),
  peak_window = c(0.5, 5),
  amplitude_min = 0.01
)

er_results

Interpreting the Results

The returned data.frame contains one row per event per channel with the following columns:

Column Description
event_index Sequential index of the event
event_onset Event onset time in seconds
channel Channel label
scr_present Whether a valid SCR was detected
scr_amplitude SCR amplitude (peak minus onset value)
scr_latency Latency from event to SCR onset
scr_rise_time Time from SCR onset to peak
scr_recovery_time Time for 50% recovery from peak

Working with ER-SCR Results

Response Rate

The SCR response rate is an important measure that indicates the proportion of trials that elicited a detectable SCR.

# Overall response rate
response_rate <- mean(er_results$scr_present)
cat("Response rate:", round(response_rate * 100, 1), "%\n")

Mean Amplitude

When computing mean amplitudes, a common approach is to include only trials with a valid SCR (responders-only) or to set non-responder amplitudes to zero.

# Responders-only mean amplitude
responders <- er_results[er_results$scr_present, ]
if (nrow(responders) > 0) {
  mean_amp <- mean(responders$scr_amplitude, na.rm = TRUE)
  cat("Mean SCR amplitude (responders):", round(mean_amp, 3), "uS\n")
}

# All-trials mean (non-responders = 0)
er_results$amplitude_zero <- ifelse(er_results$scr_present,
                                     er_results$scr_amplitude, 0)
mean_amp_all <- mean(er_results$amplitude_zero)
cat("Mean SCR amplitude (all trials):", round(mean_amp_all, 3), "uS\n")

Comparing Conditions

When events have different types or values, you can compare SCR features across conditions.

# If events have different types, filter by event type
# Example: comparing CS+ vs CS- in a conditioning paradigm

# Get event metadata
events <- getEvents(x)
event_df <- events@events

# Merge event metadata with ER-SCR results
er_results$event_value <- event_df$value[er_results$event_index]

# Compare conditions (using base R)
by_condition <- split(er_results, er_results$event_value)
condition_summary <- do.call(rbind, lapply(names(by_condition), function(cond) {
  d <- by_condition[[cond]]
  data.frame(
    condition = cond,
    n_trials = nrow(d),
    response_rate = mean(d$scr_present),
    mean_amplitude = mean(d$scr_amplitude, na.rm = TRUE),
    mean_latency = mean(d$scr_latency, na.rm = TRUE),
    stringsAsFactors = FALSE
  )
}))
condition_summary

Plotting the Signal with Event Markers 

# Plot the full signal with event markers
plotEda(x, show_events = TRUE, main = "EDA with Stimulus Events")

# Zoom into a specific event window
plotEda(x, time_range = c(50, 80), show_events = TRUE)

Plotting SCR Peaks 

# Detect and plot all SCR peaks
peaks <- edaPeaks(x)
plotPeaks(x, peaks = peaks, channel = 1)

Data Transformation for Statistics

SCR amplitudes are typically right-skewed. Before parametric statistical tests, consider applying a log or square root transformation.

# Log transform the amplitudes
if (any(er_results$scr_present)) {
  er_results$log_amplitude <- log1p(er_results$scr_amplitude)

  # Or use the full-signal transform
  x_log <- edaTransform(x, method = "log")
}
library(PhysioEDA)

# Step 1: Create data with events
x <- make_eda_with_scr(n_time = 6000, sr = 10, n_events = 4)

# Step 2: Preprocess
x <- edaFilter(x, type = "lowpass", cutoff = 1.0)
x <- edaArtifact(x, correct = "interpolate")

# Step 3: Decompose
x <- edaDecompose(x, method = "cda")

# Step 4: Extract event-related SCRs
er_results <- edaErscr(x, onset_window = c(1, 4), amplitude_min = 0.01)

# Step 5: Summarize
cat("Event-Related SCR Summary\n")
cat("=========================\n")
cat("Total events:", nrow(er_results), "\n")
cat("Response rate:", round(mean(er_results$scr_present) * 100, 1), "%\n")
if (any(er_results$scr_present)) {
  resp <- er_results[er_results$scr_present, ]
  cat("Mean amplitude:", round(mean(resp$scr_amplitude), 3), "uS\n")
  cat("Mean latency:", round(mean(resp$scr_latency), 2), "s\n")
  cat("Mean rise time:", round(mean(resp$scr_rise_time), 2), "s\n")
}

# Step 6: Visualize
plotDecompose(x, channel = 1)
plotEda(x, show_events = TRUE)

Best Practices

  1. Onset window: Use 1-4 seconds for typical sympathetic response latencies. Adjust for your paradigm if needed (e.g., shorter for startle responses).

  2. Amplitude threshold: The conventional minimum is 0.01-0.05 microsiemens. Use a consistent threshold across all participants and conditions.

  3. Decomposition method: CDA is recommended for event-related analyses as it provides better temporal resolution of overlapping SCRs than simple highpass filtering.

  4. Non-responders: Decide a priori whether to analyze responders only or to score non-responders as zero amplitude. Report which approach was used.

  5. Multiple channels: If recording from multiple sites, analyze each channel separately. The results data.frame includes a channel column for this purpose.

References

  • Bach, D.R., et al. (2010). Modelling event-related skin conductance responses. International Journal of Psychophysiology, 75(3), 349-356.
  • Benedek, M., & Kaernbach, C. (2010). A continuous measure of phasic electrodermal activity. Journal of Neuroscience Methods, 190(1), 80-91.
  • Boucsein, W. (2012). Electrodermal Activity. 2nd ed. Springer.

Session Info