spacr.response_distribution¶

Compare a response distribution before and after transformation.

The module applies the same transformation and distribution classifier used by the regression pipeline. Its combined histogram therefore provides a diagnostic of how the selected transformation changes the response and the candidate model family.

Functions¶

`caption`(→ str)	Format a statistical caption for a `compare()` result.
`compare`(→ Dict[str, Any])	Compare response distributions before and after transformation.
`describe`(→ Dict[str, Any])	Summarize and classify a response distribution.
`fast_panel`(values, transform[, plot, dependent_variable])	Draw the before/after comparison on a pyqtgraph plot.
`panel`(values, transform[, ax, dependent_variable])	Plot response distributions before and after transformation.
`transformed`(→ numpy.ndarray)	Apply the regression pipeline's response transformation.

Module Contents¶

spacr.response_distribution.caption(result: Dict[str, Any]) → str[source]¶

Format a statistical caption for a compare() result.

Parameters:: result – response-distribution comparison result to render.

spacr.response_distribution.compare(values: Sequence[float], transform: str) → Dict[str, Any][source]¶

Compare response distributions before and after transformation.

Parameters:

values (sequence of float) – Response values. Non-finite values are excluded.
transform (str) – Transformation to evaluate.

Returns:

dict – The original and transformed arrays, their summaries, the normalized transformation name, and flags indicating whether values changed and whether the transformed values require a separate axis.

spacr.response_distribution.describe(values: Sequence[float]) → Dict[str, Any][source]¶

Summarize and classify a response distribution.

Parameters:: values (sequence of float) – Response values before or after transformation. Non-finite values are excluded.
Returns:: dict – Sample size, range, skewness, D’Agostino normality-test p-value, regression-family identifier, and display label. The family is empty when fewer than eight finite observations are available or classification fails.

spacr.response_distribution.fast_panel(values: Sequence[float], transform: str, plot=None, dependent_variable: str = '')[source]¶

Draw the before/after comparison on a pyqtgraph plot.

The same picture panel() draws, on the screen’s own renderer, so the figure a run writes and the figure a tab shows are one scene. Both distributions go on ONE pair of axes as outlines rather than on two stacked panels: the question is whether the transform moved the shape, and two shapes on separate axes with separate scales is the one layout that cannot answer it.

Parameters:

values (sequence of float) – Response values. Non-finite values are excluded.
transform (str) – Transformation name, as transformed() accepts.
plot (FastPlot or None, default=None) – Where to draw. One is created when omitted.
dependent_variable (str, default="") – The response’s name, for the axis label.

Returns:

FastPlot or None – The plot drawn on, or None when there is nothing finite to draw.

spacr.response_distribution.panel(values: Sequence[float], transform: str, ax=None, dependent_variable: str = '')[source]¶

Plot response distributions before and after transformation.

Parameters:

values (sequence of float) – Response values to compare.
transform (str) – Transformation to apply.
ax (matplotlib.axes.Axes, optional) – Axes on which to draw. A standalone figure and axes are created when omitted.
dependent_variable (str, optional) – Response name shown on the horizontal axis.

Returns:

dict – Result from compare(), augmented with the primary axes under "axes".

spacr.response_distribution.transformed(values: Sequence[float], transform: str) → numpy.ndarray[source]¶

Apply the regression pipeline’s response transformation.

Parameters:

values (sequence of float) – Response values to transform.
transform (str) – Transformation accepted by spacr.ml.apply_transformation().

Returns:

numpy.ndarray – Transformed values. Values are returned unchanged for "none", an unsupported transformation, or a transformation that cannot be applied.

Nested helpers¶

caption.one(part: Dict[str, Any]) → str¶

Format one side of the captured before/after comparison.

Parameters:: part – distribution summary returned by describe().
Returns:: the display name alone when no family or finite statistics are available; otherwise the name followed by finite normality and skewness statistics.

spacr/response_distribution.py:160