Principles behind variance misallocation in temporal exploratory factor analysis for ERP data: Insights from an inter-factor covariance decomposition

Scharf, Florian; Nestler, Steffen

doi:10.1016/j.ijpsycho.2018.03.019

Cited by 22 publications

(56 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The biases in the factor loadings for both ESEM and EFA followed the same pattern as in previous simulations (Scharf & Nestler, ); that is, we observed differentiable biases due to the violation of orthogonality constraints (i.e.,

{φ_{12}}_{} \neq

0 or, for EFA, topographic overlap) and due to the rotation criterion. The former bias caused spurious cross‐loadings when varimax was used in spite of substantial factor correlations in the population.…”

Section: Discussionsupporting

confidence: 85%

“…Differences in the factor correlation estimates were more profound: The factor correlation was strongly inflated by EFA (

{\overset{φ}{true}}_{12} = 0.74

), whereas ESEM indeed provided much less biased factor correlation estimates (

\overset{true^}{{\overset{false^}{φ}}_{12}} = 0.23

). These results are in line with the notion that (co)variance estimates are the sum of the different sources of (co)variance in the raw data matrix (Dien, ; Scharf & Nestler, ) and demonstrate that ESEM but not EFA could separate group variance, topographic variance, and participant variance in our example data set.…”

Section: Introductionsupporting

confidence: 90%

“…The initial model is typically used to determine the number of factors to be retained (but see also Kayser & Tenke, 2003), for instance, using parallel analysis (Horn, 1965). The factor loadings of the extracted factors are rotated utilizing an oblique rotation procedure (Dien, 1998(Dien, , 2010Scharf & Nestler, 2018) yielding the final (rotated) estimates of the factor loadings and the factor correlations.…”

Section: Efamentioning

confidence: 99%

“…Previous research has mainly focused on variance misallocation due to biased factor loading estimates. This may occur either as a consequence of implausibly choosing an orthogonal rotation (orthogonality bias; Dien, ; Dien et al, ; Scharf & Nestler, ) or as a consequence of the rotation itself, which biases the estimates toward its simplicity criterion (rotation bias; Dien, ; Dien et al, ; Möcks & Verleger, ; Scharf & Nestler, ). The rotation bias is a function of the temporal overlap of the factors and gets stronger the more two factors load on the same sampling points.…”

Section: Introductionmentioning

confidence: 99%

“…Hence, when error variance approaches zero and/or the number of observed variables is high, the differences between EFA and PCA become negligible (Widaman, 2007). It has been noted that these conditions typically hold for ERP data (Dien, Beal, & Berg, 2005;Scharf & Nestler, 2018). However, it is unknown as to how far this is a general condition that always holds for applications of the method.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Exploratory structural equation modeling for event‐related potential data—An all‐in‐one approach?

Scharf

Nestler

2018

Psychophysiology

Self Cite

View full text Add to dashboard Cite

ERP data are characterized by high dimensionality and a mixture of constituting signals and are thus challenging for researchers to analyze. To address these challenges, exploratory factor analysis (EFA) has been used to provide estimates of the unobserved factors and to use these estimates for further statistical analyses (e.g., analyses of group effects). However, the EFA approach is prone to biases due to assigning individual factor scores to each observation as an intermediate step and does not properly consider participants, electrodes, and groups/conditions as differentiable sources of factor variance, with the consequence that factor correlations are inaccurately estimated. Here, we suggest exploratory structural equation modeling (ESEM) as a potential approach to address these limitations. ESEM may handle the complexity of ERP data more appropriately because multiple sources of variance can be formally taken into consideration. We demonstrate the application of ESEM to ERP data (in comparison with EFA) with an illustrative example and report the results of a small simulation study in which ESEM clearly outperformed EFA with respect to accurate estimation of the population factor loadings, population factor correlations, and group differences. We discuss how robust statistical inference can be conducted within the ESEM approach. We conclude that ESEM naturally extends the current EFA approach for ERP data and that it can provide a coherent and flexible analysis framework for all kinds of ERP research questions.

show abstract

{φ_{12}}_{} \neq

Section: Discussionsupporting

confidence: 85%

“…Differences in the factor correlation estimates were more profound: The factor correlation was strongly inflated by EFA (

{\overset{φ}{true}}_{12} = 0.74

), whereas ESEM indeed provided much less biased factor correlation estimates (

\overset{true^}{{\overset{false^}{φ}}_{12}} = 0.23

Section: Introductionsupporting

confidence: 90%

Section: Efamentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Exploratory structural equation modeling for event‐related potential data—An all‐in‐one approach?

Scharf

Nestler

2018

Psychophysiology

Self Cite

View full text Add to dashboard Cite

show abstract

Somatosensory omissions reveal action‐related predictive processing

Dercksen,

Widmann,

Noesselt

et al. 2023

Human Brain Mapping

View full text Add to dashboard Cite

The intricate relation between action and somatosensory perception has been studied extensively in the past decades. Generally, a forward model is thought to predict the somatosensory consequences of an action. These models propose that when an action is reliably coupled to a tactile stimulus, unexpected absence of the stimulus should elicit prediction error. Although such omission responses have been demonstrated in the auditory modality, it remains unknown whether this mechanism generalizes across modalities. This study therefore aimed to record action‐induced somatosensory omission responses using EEG in humans. Self‐paced button presses were coupled to somatosensory stimuli in 88% of trials, allowing a prediction, or in 50% of trials, not allowing a prediction. In the 88% condition, stimulus omission resulted in a neural response consisting of multiple components, as revealed by temporal principal component analysis. The oN1 response suggests similar sensory sources as stimulus‐evoked activity, but an origin outside primary cortex. Subsequent oN2 and oP3 responses, as previously observed in the auditory domain, likely reflect modality‐unspecific higher order processes. Together, findings straightforwardly demonstrate somatosensory predictions during action and provide evidence for a partially amodal mechanism of prediction error generation.

show abstract

The oddball effect on P3 disappears when feature relevance or feature-response mappings are unknown

et al. 2018

View full text Add to dashboard Cite

The P3b component of human event-related EEG potentials is larger with rare than frequent task-relevant stimuli. In a previous study, this oddball effect was much reduced when stimulus-response (S-R) mappings were still undefined at stimulus presentation (being later provided by response prompts). This reduction may reflect P3b's dependence on transmitted information which might be any relevant information (informational value hypothesis) or, more specifically, information about how to respond (S-R link hypothesis). To distinguish between these two hypotheses and clarify their differences from classical stimulus evaluation hypothesis, we added a second dimension by presenting colored letters, with both colors and letters varying between a rare and a frequent alternative. Response prompts, presented half a second later, were, in different blocks, constant or variable across trials with respect to S-R mapping and with respect to the relevant dimension (color or letter). With partial information, when only one of these two factors is known at stimulus presentation (by being constant across trials), the hypotheses differ in their predictions. The oddball effect will be abolished according to S-R link hypothesis because knowledge of both factors is needed to determine the response, but will only be reduced according to informational value hypothesis and be fully maintained according to stimulus evaluation hypothesis. In fact, oddball effects only occurred with knowledge of both factors, i.e., if both the relevant dimension and its mapping to responses were constant across trials. These results confirm the preeminent role of knowledge about responses for eliciting P3.

show abstract

Principles behind variance misallocation in temporal exploratory factor analysis for ERP data: Insights from an inter-factor covariance decomposition

Cited by 22 publications

References 56 publications

Exploratory structural equation modeling for event‐related potential data—An all‐in‐one approach?

Exploratory structural equation modeling for event‐related potential data—An all‐in‐one approach?

Somatosensory omissions reveal action‐related predictive processing

The oddball effect on P3 disappears when feature relevance or feature-response mappings are unknown

Contact Info

Product

Resources

About