Interpretability, not statistical criteria alone, should decide how many components to retain

Evidence

Eigenvalues and parallel analyses are useful methods for determining the number of components to retain, but foremost in this process is the interpretability of components (DeVellis, 2016). For this reason, as the items on the second component were separated from similar items on the first component, a one-component model was retained for the Societal scale. — Naheed Ahmed, Measurement of Perceived Interpersonal…, 70

Categorization

• XREF: Pairs with standard factor-retention debates (Kaiser criterion, parallel analysis, scree tests) familiar from psychometrics coursework.
• DEFINE: Clarifies that eigenvalue and parallel analysis are retention heuristics subordinate to substantive interpretability.