Returns the quality of representation of each group on each component.

Map projected coordinates back to quaternions on the manifold. Parameters ---------- X Projected coordinates from

Beers dataset. The data is taken from https://github.com/philipperemy/beer-dataset.

Burgundy wines dataset. Source: https://personal.utdallas.edu/~herve/Abdi-MCA2007-pretty.pdf

The Decathlon dataset from FactoMineR.

Per capita energy mix by country in 2019. Each row corresponds to a country. There is one column for each energy source. A value corresponds

Voting data for the 2022 French elections, by region. The [original dataset](https://www.data.gouv.fr/fr/datasets/resultats-du-premier-tour-de-le

Hearthstone standard cards. Source: https://gist.github.com/MaxHalford/32ed2c80672d7391ec5b4e6f291f14c1

The Poison dataset from FactoMineR. Survey of 55 children who fell sick during a school meal. The first two columns are numeric (``Age``, ``T

Premier League dataset. The data is taken from Wikipedia, using pd.read_html.

Punctuation marks of six French writers.

Robot IMU quaternion rotations on different surfaces. Each row is a unit quaternion (qw, qx, qy, qz) in scalar-first convention representing

Logarithmic map: project manifold points to tangent space at base. Parameters ---------- base The base point on t

Returns the contribution of each group's PCA axes to each global MFA component.

Returns the correlations between each group's PCA axes and the global MFA axes.

Returns the percentage of explained inertia per principal component.

Returns the percentage of explained inertia per principal component.

Plot the projected data in the tangent space. Parameters ---------- X Rotation data (same format as fit).

Plot the partial axes correlations. Each arrow represents a group's PCA component projected onto the global MFA space. Arrows close t

Returns the final reference shape.

Returns the row principal coordinates. Equivalent to transform().

Returns the cosine similarities between the rows and their principal components. The row cosine similarities are obtained by calculating the

Squared cosine similarities (cos2) of each row on each component.

Returns the row standard coordinates. The row standard coordinates are obtained by dividing each row principal coordinate by it's ass

Small synthetic quaternion dataset with known structure.

Dimension of the tangent space.

>>> wines = prince.datasets.load_burgundy_wines() >>> wines = wines.drop(columns=["Oak type"], level=0) >>> mca = prince.MCA(n_componen

Figure 4: category-level coordinates (barycentres of row coordinates). FactoMineR: famd$quali.var$coord Each category's coordinate is

Figure 2: relationship square (r² for quanti, η² for quali). FactoMineR: famd$var$coord.

Column outputs use a clean 2-level (group, variable) MultiIndex (issue #242). Categorical indicator labels read ``variable__category`` rather

Reproduce issue #242: clean (group, variable) MultiIndex labels for a mixed MFA.

The X-taking column_coordinates method is not implemented yet (see #242).

Figure 3: correlation circle for quantitative variables. FactoMineR: famd$quanti.var$coord Signed correlations between standardized q

The first component should align with the z-axis (largest variance).

In our specific case of all-same-shape circles, the shapes should align perfectly.

Aligning a single shape should return the same shape, just normalized.