Skip to content

Plots

plots

Plots for qlPCA: scree plot, transformation plots, biplot (matplotlib, imported lazily).

Style: light surface, thin lines, recessive grid, blue for the main series and orange for the comparison series; text stays in neutral ink.

scree_plot

scree_plot(eigenvalues, ax=None, label=None, reference=None, reference_label=None, kaiser=True, title=None)

Scree plot of the eigenvalues of the correlation matrix of the (transformed) variables.

Solutions of different dimension are not nested, so compare the scree plots of p - 1, p and p + 1.

Parameters:

Name Type Description Default
eigenvalues array - like or QLPCA

Eigenvalues, or a fitted model (its all_eigenvalues_ are used).

required
ax matplotlib Axes

Axes to draw on.

None
label str

Legend label of the series.

None
reference array - like or QLPCA

A second series, for example the PCA eigenvalues.

None
reference_label str

Legend label of the reference series.

None
kaiser bool

Draw the line eigenvalue = 1.

True
title str

Axes title.

None

Returns:

Type Description
matplotlib Axes

transformation_plots

transformation_plots(model, X, names=None, ncols=4, figsize=None, n_points=200)

Small multiples of the optimal transformations: f_j against the original variable.

Points are the observations, the line is the fitted spline, dashed verticals are the interior knots.

Parameters:

Name Type Description Default
model QLPCA

A fitted model.

required
X array-like of shape (n_samples, n_features)

The data the model was fitted on.

required
names list of str

Variable names (default: feature_names_in_ or x1, x2, ...).

None
ncols int

Number of panel columns.

4
figsize tuple

Figure size.

None
n_points int

Grid points per curve.

200

Returns:

Type Description
matplotlib Figure

biplot

biplot(model, X=None, names=None, labels=None, components=(0, 1), ax=None, arrow_scale=None)

Biplot: object scores as points and loadings as arrows.

Arrows are rescaled by one common factor so that all tips lie within the range of the scores. The loadings refer to the transformed variables; a variable with a V-shaped transformation has an arbitrary sign (DESIGN D6).

Parameters:

Name Type Description Default
model QLPCA

A fitted model.

required
X array - like

Not used (the training scores are drawn); kept for a uniform signature.

None
names list of str

Variable names for the arrows.

None
labels list of str

Labels of the points.

None
components tuple of int

Components on the horizontal and vertical axes.

(0, 1)
ax matplotlib Axes

Axes to draw on.

None
arrow_scale float

Factor applied to the loadings (default: chosen automatically).

None

Returns:

Type Description
matplotlib Axes

set_style

set_style()

Set the matplotlib defaults used by the notebooks (white surface, neutral ink, thin lines).

algorithm_diagram

algorithm_diagram(ax=None)

Flow diagram of the qlPCA algorithm (thesis Figure 3.1), with the ALS loop shaded.

Parameters:

Name Type Description Default
ax matplotlib Axes

Axes to draw on.

None

Returns:

Type Description
matplotlib Axes