SingleMetricBoxplot

class gaitmap_challenges.visualization.SingleMetricBoxplot(cv_results: Dict[str, DataFrame], metric: str, use_aggregation: Literal['fold', 'single'] = 'single', overlay_scatter: bool = True, force_order: Optional[Sequence[str]] = None, label_grouper: Optional[Callable[[Series], Categorical]] = None, invert_grouping: bool = False, matplotlib_boxplot_props: Optional[Dict[str, Any]] = None)

Create a boxplot for a single metric.

The plot can be either created with matplotlib or bokeh, by using the respective methods.

Most parameters are shared between both plotting backends. Specific parameters are prefixed with bokeh_ or matplotlib_ or are expected to be passed to the method directly.

Parameters:

cv_results

A list of cv results of multiple algorithms as loaded by load_run.

metric

The metric to plot. This should be the error metric without the “test_” and/or “single_” prefix.

use_aggregation

Whether to use the “fold” or “single” aggregation. With “fold” each point in the final plot is a single fold. With “single” we pool all datapoint values from all test folds, so that each point in the final plot is one datapoint (i.e. one participant) In case of single, we look for the metric in the “test_single_<metric>” column. In case of fold, we look for the metric in the “test_<metric>” column. This further changes what other parameters make sense.

overlay_scatter

Whether to overlay a scatterplot on top of the boxplot.

label_grouper

A function that returns a group label for each data point. This can be used to split the data into multiple groups that are plotted independently to effectively create grouped boxplots. Most likely you want to use group_by_data_label to create this function. Note, that this setting only makes sense when use_aggregation="single".

invert_grouping

Whether to invert the grouping returned by label_grouper. This will just change the order of the boxplots and which boxplots are grouped together. By default, all groups of one algorithm are plotted next to each other. With invert_grouping=True all boxplots of one group are plotted next to each other.

force_order

A list of algorithm names that specifies the order in which the boxplots are plotted. If you want to force the order of your groups, fix the order in the label_grouper function.

matplotlib_boxplot_props

Additional properties to pass to the sns.boxplot function in the matplotlib backend.

Attributes:

force_order

label_grouper

matplotlib_boxplot_props

Methods

bokeh()	Create the plot using bokeh.
matplotlib([ax])	Create the plot using matplotlib.

__init__(cv_results: Dict[str, DataFrame], metric: str, use_aggregation: Literal['fold', 'single'] = 'single', overlay_scatter: bool = True, force_order: Optional[Sequence[str]] = None, label_grouper: Optional[Callable[[Series], Categorical]] = None, invert_grouping: bool = False, matplotlib_boxplot_props: Optional[Dict[str, Any]] = None) → None

bokeh()

Create the plot using bokeh.

This creates a plot object that can be displayed using bokeh.plotting.show.

matplotlib(ax: Optional[Axes] = None)

Create the plot using matplotlib.

You can optionally pass an existing matplotlib axes object to plot into.