Lecture 4: Exploratory Factor Analysis (Alternate) Flashcards
(58 cards)
What is Exploratory Factor Analysis (EFA)?
A statistical technique used to identify latent psychological constructs based on patterns of shared variance among measured variables.
Why do researchers use observable indicators like questionnaire responses in EFA?
Because unobservable mental processes (e.g., intelligence, anxiety) are inferred via observable indicators that may imperfectly reflect the constructs of interest.
What does the core premise of EFA state about shared variance?
Shared variance among variables reflects a smaller number of latent factors.
How are latent factors inferred in EFA?
Through their pattern of associations with the variables.
What do coherent clusters of strongly correlated variables represent in EFA?
Underlying constructs or “factors.”
What does a factor loading quantify in EFA?
The strength of the association between each variable and a given factor.
What types of research questions does EFA help address?
1) What constructs are being measured by a set of variables? 2) What is the underlying structure of a construct? Is it unidimensional or multidimensional? 3) Are two constructs empirically distinct or overlapping?
What is the goal of EFA?
Arrive at the most parsimonious factor structure.
What are the two major steps of the EFA process?
Factor extraction and factor rotation.
What does the extraction step in EFA identify?
The minimum number of dimensions (factors) that can account for the shared variance in a dataset.
How does factor rotation enhance interpretability in EFA?
By maximizing high loadings and minimizing low ones without altering the total variance explained.
What are common reasons researchers rarely retain all extracted factors in EFA?
Some pick up only noise, explain little variance, or include only a single item.
What is Kaiser’s Criterion in EFA?
Retain factors with eigenvalues > 1, as they explain more variance than a single variable.
What does the Scree Test suggest about factor retention?
Retain factors in the steep, descending part of an eigenvalue plot, discarding those in the flat “scree” region.
What is the principle behind Parallel Analysis?
Retain only factors whose eigenvalues are larger than those from randomly generated datasets.
How does Velicer’s Minimum Average Partial (MAP) Test work?
Sequentially partials out factors and retains the number that minimizes average squared partial correlations.
Why is it best practice to use multiple methods for deciding the number of factors to retain?
Because each method has strengths and limitations, and different methods may yield conflicting conclusions.
What are two major limitations of factor extraction in EFA?
It doesn’t define well which variables contribute to each factor and late factors operate under more constraints.
What does a complex factor structure indicate in EFA?
A structure in which a factor consists of variables with both high and low loadings, making interpretation difficult.
What is the minimum item-to-factor ratio recommended during EFA planning?
Minimum of 3, ideally 5–6 items per factor.
What is the recommended sample size for EFA?
Generally more than 50 participants and ideally at least 5 times the number of variables.
What are key assumptions to check before conducting EFA?
Interval scale measurement, sufficient item variance, linear correlations between variables, and approximate normality of score distributions.
What is a key limitation of Principal Components Analysis (PCA) for psychology?
It assumes all variance is shared (no error or unique variance) and often overestimates factor loadings.
How does Principal Axis Factoring (PAF) estimate communalities?
From the empirical correlation matrix using values less than 1.
