ROTATION OF FACTORS
The different methods of factor analysis first extract a set a factors from a data set. These factors
are almost always orthogonal and are ordered according to the proportion of the variance of the
original data that these factors explain. In general, only a (small) subset of factors is kept for
further consideration and the remaining factors are considered as either irrelevant or nonexistent
(i.e., they are assumed to reflect measurement error or noise). In order to make the interpretation
of the factors that are considered relevant, the first selection step is generally followed by a
rotation of the factors that were retained. Two main types of rotation are used: orthogonal when
the new axes are also orthogonal to each other, and oblique when the new axes are not required
to be orthogonal to each other. Because the rotations are always performed in a subspace (the so-
called factor space), the new axes will always explain less variance than the original factors
(which are computed to be optimal), but obviously the part of variance explained by the total
subspace after rotation is the same as it was before rotation (only the partition of the variance has
changed). Because the rotated axes are not defined according to a statistical criterion, their most
important reason is to facilitate the interpretation.
It is important to stress that because the rotations always take place in a subspace (i.e., the space
of the retained factors), the choice of this subspace strongly influences the result of the rotation.
Therefore, in the practice of rotation in factor analysis, it is strongly recommended to try several
sizes for the subspace of the retained factors in order to assess the robustness of the interpretation
of the rotation.
ROTATING: WHEN AND WHY?
Most of the rationale for rotating factors comes from Thurstone (1947) and Cattell (1978) who
defended its use because this procedure simplifies the factor structure and therefore makes its
interpretation easier and more reliable (i.e., easier to replicate with different data samples).
Thurstone suggested five criteria to identify a simple structure. According to these criteria, still
often reported in the literature, a matrix of loadings (where the rows correspond to the original
variables and the columns to the factors) is simple if: 1. each row contains at least one zero; 2.
for each column, there are at least as many zeros as there are columns (i.e., number of factors
kept); 3. for any pair of factors, there are some variables with zero loadings on one factor and
large loadings on the other factor; 4. for any pair of factors, there is a sizable proportion of zero
loadings; 5. for any pair of factors, there is only a small number of large loadings. Rotations of
factors can (and used to) be done graphically, but are mostly obtained analytically and
necessitate to specify mathematically the notion of simple structure in order to implement it with
a computer program.
TYPES OF ROTATION
Two main types of rotations are called-
• Orthogonal
• Oblique
• Orthogonal rotation An orthogonal rotation is specified by a rotation matrix
denoted R, where the rows stand for the original factors and the columns for the new (rotated)
factors.
Orthogonal rotations maiantain the independence of factors, i.e., the angles between the axes are
kept at 90 degrees. For example there are two factors and we rotate factors I and II
orthogonally , we swing both axes together, maintaining the right angle between them. This
means that the correlation between the factors is zero
(you can draw here fig 37.5 given on page 674 in your notes)
TYPES OF ORTHOGONAL ROTATION
• Varimax: Varimax, which was developed by Kaiser (1958), is indubitably the most
popular rotation method by far. For varimax a simple solution means that each factor has a small
number of large loadings and a large number of zero (or small) loadings. This simplifies the
interpretation because, after a varimax rotation, each original variable tends to be associated with
one (or a small number) of factors, and each factor represents only a small number of variables.
In addition, the factors can often be interpreted from the opposition of few variables with
positive loadings to few variables with negative loadings. Formally varimax searches for a
rotation (i.e., a linear combination) of the original factors such that the variance of the loadings is
maximized, which amounts to maximizing
• Quartimax Method: A rotation method that minimizes the number of factors
needed to explain each variable. This method simplifies the interpretation of the observed
variables.
• Equamax Method: A rotation method that is a combination of the varimax
method, which simplifies the factors, and the quartimax method, which simplifies the variables.
The number of variables that load highly on a factor and the number of factors needed to explain
a variable are minimized.
Other methods exist, but none approaches varimax in popularity
• Oblique Rotations
Obliqueness means that the factors are correlated. In oblique rotations the new axes are free to
take any position in the factor space, but the degree of correlation allowed among factors is, in
general, small because two highly correlated factors are better interpreted as only one factor.
Oblique rotations, therefore, relax the orthogonality constraint in order to gain simplicity in the
interpretation. They were strongly recommended by Thurstone, but are used more rarely than
their orthogonal counterparts.
(you can add here fig 37.6 given on page 675 of your notes)
TYPES OF OBLIQUE ROTATION
• Direct Oblimin Method: A method for oblique (nonorthogonal) rotation. When
delta equals 0 (the default), solutions are most oblique. As delta becomes more negative, the
factors become less oblique.
• Promax Rotation. An oblique rotation, which allows factors to be correlated. This
rotation can be calculated more quickly than a direct oblimin rotation, so it is useful for large
datasets.
For oblique rotations, the promax rotation has the advantage of being fast and conceptually
simple. Its name derives from procrustean rotation because it tries to fit a target matrix which has
a simple structure.
Even though, in principle, the results of oblique rotations could be presented graphically they are
almost always interpreted by looking at the correlations between the rotated axis and the original
variables. These correlations are interpreted as loadings.
Oblique rotations are much less popular than their orthogonal counterparts, but this trend may
change as new techniques, such as independent component analysis, are developed. This recent
technique, originally created in the domain of signal processing and neural networks, derives
directly from the data matrix an oblique solution that maximizes statistical independence.