performs a (classical) Correspondence Analysis on X, a non-negative
matrix with no zero row or column totals.
Returns two matrices of row and column scores for numberAxes eigenvalues.
All options passed to scoresOptionD are passed on to scores.
Note that, in this version of CA, the trivial (unit singular value)
solution is implicitly removed from the analysis (see Gower and Hand
1996, p.180) by scaling the input matrix to unit total.
performs a (classical) Correspondence Analysis on X, a non-negative
matrix with no zero row or column totals.
Returns two matrices of row and column scores for numberAxes eigenvalues.
All options passed to scoresOptionD are passed on to scores.
Note that, in this version of CA, the trivial (unit singular value)
solution is implicitly removed from the analysis (see Gower and Hand
1996, p.180) by scaling the input matrix to unit total.
returns scores from row and column Principal Axes (Rs,`Cs`) and
the Singular Values (S).
The scaling parameters specifies which scaling to use for the scores:
SCALING_PRINCIPAL: simultaneous non-least-squares approximation to
chi-square distances of rows and columns (also known as "symmetric"
scores); compromise between the two former methods
If transition is set to CA_TRANSITION_COLUMNTOROW, Row_PA is expected
to be in a form so that post-multiplication with the column scores will
result in the row scores; similarly, if transition is set to
CA_TRANSITION_ROWTOCOLUMN, Col_PA is expected to be in a form that
post-multiplication with the row scores will result in the column scores.
This uses the transition formulae of CA (Gower and Hand 1996, p. 181).
If annotate is true, the resulting row and column coordinattes are
printed to stdout.
If numberAxes is not None, only the first numberAxes of the
Principal Axes/Singular Values are used for computing the coordinates.
analysis of a statistical triplet (X,C,R) (Escoufier 1987)
or Generalized Singular Value Decomposition (Greenacre 1984).
Parameters:
X: n x p data matrix (possibly after centering/standardizing)
C: matrix used for calculating the column distances (column metrics)
R: matrix of row weights (row metrics)
scoresOptionD: additional options to be passed to scores
For standard PCA, we X is centered or standardized, C is a p x p
identity matrix, and R is 1/n times a n x n identity matrix.
The function performs a standard SVD of R**1/2*X*C**1/2.
Note that R and C have to be symmetric for the present implementation
Value:
3-tuple of arrays ::
(<row scores>, <singular values>, <column scores>)