Decide artefact thresholds and average trials
Review threshold decisions
Select Decide statistical artefact thresholds and Review decisions to inspect the prepared statistics. Each parameter heatmap has channels on one axis and trials on the other. Sorting helps identify outliers; coloured marks denote channel/trial decisions, not necessarily rejection of every marked trial.
- Inspect the active parameters and their channel thresholds.
- Use Calc approx. to calculate trial approximation errors.
- Inspect the resulting distribution and set the acceptable trial-error limit.
- Inspect representative trials when needed, then accept the decision.
The approximation error assesses the consequences of reconstructing unavailable channels. It is not the same quantity as a raw amplitude threshold. Changing channel decisions requires updating that calculation.
A zero median approximation error can occur when most trials need no repair; it does not mean that every trial is invalid. Zero-error trials are not rejected merely for equalling a zero limit. If all trials are rejected, investigate the parameters, channel configuration and saved decisions rather than loosening thresholds only to obtain an average.
Unattended thresholding
Unattended uses configured automatic thresholds or compatible saved decisions. It still performs the numerical threshold/approximation work and writes decisions; it skips interactive inspection and heatmap preparation.
Threshold decisions are stored in .WE files, often .SES.AR.WE, with
human-readable editing details in .WE.TXT. An existing decision may be reused
when reuse is enabled. If an unexpected result persists, check whether the log
says that a saved WE file was reused.
Finish averaging and results
The final step reads the epochs and decisions, repairs eligible channel/trial data, applies the selected reference and averages accepted trials by condition. The log reports retained trial counts and channel-quality summaries.
Optional event-group profiles can merge conditions or constrain accepted trials before averaging.
Typical products are:
| File | Meaning |
|---|---|
*.at<condition>.ar |
A condition average |
*.APP<condition> |
Optional clean/approximated trials for one condition |
*.APP.SES |
Global trial container used with its decisions/condition provenance |
*.trial-selection.json |
Exact selected trials for a non-default grouping profile |
The global APP.SES retains original trial positions and rejected-trial placeholders; it is not a plain list in which every entry is an accepted trial. Downstream WAVELET uses the accompanying provenance to identify usable trials.
EMEGS checks output writes. A disk-full or permission error means the stage has not completed, even when some earlier files exist. Use Processing Outcomes and Retained Trials to inspect completeness and selected-profile counts.