Statistics Glossary → Factor rotation (varimax, promax, oblimin)
Factor rotation (varimax, promax, oblimin)concept
A transformation in factor analysis or principal components that turns the factor axes into an easier-to-interpret 'simple structure' without changing the fit of the model.
In the unrotated solution most items load on the first factor and the factors are hard to name. Rotation turns the axes to a position in which each item loads highly on a few factors and weakly on the rest. The total variance explained and the items' communalities stay the same; only how that variance is shared among the factors changes. Kaiser normalization scales the loadings by the communalities before rotation, so that every item takes part with equal weight.
Orthogonal rotations (varimax, quartimax, equamax) keep the factors uncorrelated. Oblique rotations (promax, oblimin, geomin) allow the factors to correlate; the pattern matrix (unique contributions), the structure matrix (correlations) and the factor correlation matrix are then reported separately. Because psychological constructs are often related, oblique rotation is usually more realistic; if the factor correlations turn out small, the two approaches give similar results.
An everyday example
In a personality questionnaire, 12 items split into two factors. Before rotation every item loads between 0.5 and 0.7 on the first factor and no separation is visible. After promax rotation, six items load highly on an 'extraversion' factor and six on an 'anxiety' factor, and the correlation between the two factors is −0.25.
In MerQur
'Rotation (None / Varimax / Promax (m = 4))' in Principal Component Analysis; the 'Rotation' (Varimax, Quartimax, Equamax, Oblimin, Promax, Geomin) and 'Kaiser normalization (in rotation)' options in Exploratory Factor Analysis; and the orthoblique rotation of the divisive method in Variable Clustering.
Analyses where this concept appears
Variable Clustering (VARCLUS)Principal Component Analysis (PCA)Exploratory Factor Analysis (EFA)