corr_plot ¶
corr_plot(
data: PlotData,
show_legend: bool = True,
flip: bool = True,
threshold: float | None = None,
)
Build a correlation matrix plot of tiles, points and/or labels.
Add layers with tiles(), points() and labels(), pick a palette, then
call build(). Each layer covers the "upper", "lower" or "full"
matrix; unset types and diagonals are chosen from the combination of layers,
as lets_plot does (tiles below, points or labels above the diagonal).
Parameters:
-
data(DataFrame or dict) –Data (Pearson correlations of every numeric column pair are computed, over the rows where both are present) or a correlation matrix: square, float columns, every value in -1..1 or NaN.
-
show_legend(bool, default:True) –False hides the colorbar.
-
flip(bool, default:True) –True draws the first variable at the top of the y axis.
-
threshold(float, default:None) –Cells whose absolute correlation is below this (0..1) are left out; variables left without any cell are dropped.
Examples:
>>> import numpy as np
>>> from ggrust.bistro.corr import corr_plot
>>> rng = np.random.default_rng(42)
>>> data = {name: rng.normal(size=10) for name in "abcd"}
>>> plot = corr_plot(data).tiles().labels().palette_RdBu().build()
Points sized by the correlation, with the values on top:
>>> plot = corr_plot(data).points().labels().palette_PiYG().build()
Only the lower triangle, without the weakest correlations:
>>> plot = corr_plot(data, threshold=0.2).tiles(type="lower").labels(type="lower").build()
build ¶
build() -> PlotSpec
Create the plot.
Returns:
-
PlotSpec–An ordinary plot (layers, scales, theme,
coord_fixedandggsize); more features can be added with+.
Raises:
-
SpecError–If no layer was added (call
tiles(),points()orlabels()).
labels ¶
labels(
type: Literal | None = None,
diag: bool | None = None,
map_size: bool | None = None,
color: str | None = None,
halo_width: float | None = None,
halo_color: str | None = None,
) -> corr_plot
Write the correlations (two decimals) in the cells.
Parameters:
-
type((upper, lower, full), default:"upper") –Cells to label. Default: from the other layers.
-
diag(bool, default:None) –Whether the main diagonal is labelled. Default: only for
"full". -
map_size(bool, default:None) –True sizes the text by the absolute correlation. Default: True when the labels sit on points.
-
color(str, default:None) –Text color. Default: white on tiles or points, else colored by correlation.
-
halo_width(float, default:None) –Width of a text halo; not supported yet (ggrust warns).
-
halo_color(str, default:None) –Color of the text halo; not supported yet (ggrust warns).
Returns:
-
corr_plot–A new builder with the layer.
Raises:
-
SpecError–If
typeis not one of the accepted values.
palette_BrBG ¶
palette_BrBG() -> corr_plot
palette_PRGn ¶
palette_PRGn() -> corr_plot
palette_PiYG ¶
palette_PiYG() -> corr_plot
palette_PuOr ¶
palette_PuOr() -> corr_plot
palette_RdBu ¶
palette_RdBu() -> corr_plot
palette_RdGy ¶
palette_RdGy() -> corr_plot
palette_RdYlBu ¶
palette_RdYlBu() -> corr_plot
palette_RdYlGn ¶
palette_RdYlGn() -> corr_plot
palette_Spectral ¶
palette_Spectral() -> corr_plot
palette_gradient ¶
palette_gradient(
low: str, mid: str, high: str
) -> corr_plot
Color by a diverging gradient from low (-1) through mid (0) to high (1).
Parameters:
-
low(str) –Color of correlation -1.
-
mid(str) –Color of correlation 0.
-
high(str) –Color of correlation 1.
Returns:
-
corr_plot–A new builder with the palette.
points ¶
points(
type: Literal | None = None, diag: bool | None = None
) -> corr_plot
Draw the matrix as circles sized by absolute and colored by correlation.
Parameters:
-
type((upper, lower, full), default:"upper") –Cells to draw. Default: from the other layers.
-
diag(bool, default:None) –Whether the main diagonal is drawn. Default: only for
"full".
Returns:
-
corr_plot–A new builder with the layer.
Raises:
-
SpecError–If
typeis not one of the accepted values.
tiles ¶
tiles(
type: Literal | None = None, diag: bool | None = None
) -> corr_plot
Draw the matrix as square tiles filled by correlation.
Parameters:
-
type((upper, lower, full), default:"upper") –Cells to draw. Default: from the other layers.
-
diag(bool, default:None) –Whether the main diagonal is drawn. Default: only for
"full".
Returns:
-
corr_plot–A new builder with the layer.
Raises:
-
SpecError–If
typeis not one of the accepted values.
Points sized by the correlation, with the values on top
Only the lower triangle, without the weakest correlations