Fig 1.
Simulated data for two subsets of a typical posterior waveshape (black and red lines), reflecting a P3 component.
Both, peak latencies and 50% fractional area latencies were extracted. Peak latencies denote a single moment of maximal voltage within the designated measurement window. Fractional area latency denotes the moment when the cumulative area under the curve exceeds a defined criterion, here 50% of the overall area. Despite the high similarity of morphology across the two subsets, peak latencies would show an effect of more than 50 ms due to high frequency noise (solid vertical lines), that did not affect fractional area latency to the same amount (dashed vertical lines).
Fig 2.
Grand averages in the 1/5 data split (5 black lines) superposed by the grand-grand average across all subsets (red).
The data demonstrate that there was almost no variance across splits.
Fig 3.
Spearman-Brown corrected reliability (minimum to maximum) for the five ERP components and the four measures that provided acceptable output for all components.
The four splits were assigned to the number of trials possible, which is defined by the truncated division of the overall number of trials (132) by the size of the split (1/5: 26; 1/4: 33; 1/3: 44; 1/2: 66).
Fig 4.
Cronbach’s Alpha in the smallest split for the original data set and the dataset pruned with SME.
Table 1.
Outcome of the mixed-design ANOVAS for peak latencies and fractional area latencies in individual averages and in jackknifed data.
Data were reanalyzed after method specific reduction of the dataset based on the Standardized Measurement Error (pruned data = “pr_”). Additionally, analyses were run based on jackknife averages separately for the three age groups (cell-based = “c_”). Number of valid subjects in the analyses are given in brackets for all measures for the first component but apply to all components analyzed.
Fig 5.
Grand averages for the 5 time on task segments (1: Black bold, 2: Black thin, 3: Black dotted, 4: Red thin, 5: Red bold).
For N1, N2 and P2 components, latencies increase clearly with time on task (ToT). For the P1 only a modulation in the offset is visible that might be already driven by the following N1 effect. P3 changes before most in amplitude and morphology.
Fig 6.
Means and shaded 95% confidence intervals for local peaks (left column), 50% fractional area latency (middle column) and 50% jackknife fractional area latency (right column) for the 5 time on task segments.
Data are superposed for younger (black) and older (red) participants. Note that the time courses both of within-subject as well as of between-subject effects are highly comparable between measures.