<a id="api-esmvaltool-diag-scripts-monitor-multi-datasets"></a>

# Monitoring diagnostic to show multiple datasets in one plot (incl. biases)

Monitoring diagnostic to visualize multiple datasets.

## Description

This diagnostic can be used to visualize multiple datasets in a single plot. A
large variety of different plot types is supported.

## Currently supported plot types

Plot types can be specified with the [recipe option](#diag-monitor-multi-datasets-recipe-options) `plots`. All plots can be
customized with additional [plot-specific options](#diag-monitor-multi-datasets-plot-options).

- `1d_profile` (1D plot): plot 1D profile. Input data needs single dimension
  altitude or air_pressure.
- `annual_cycle` (1D plot): plot annual cycle. Input data needs single
  dimension month_number.
- `benchmarking_annual_cycle` (1D plot): benchmarking version of
  `annual_cycle`. Input data needs single dimension month_number.
- `benchmarking_boxplot` (boxplot): plot boxplot. Input data needs to be
  scalar.
- `benchmarking_diurnal_cycle` (1D plot): benchmarking version of
  `diurnal_cycle`. Input data needs single dimension hour.
- `benchmarking_map` (2D plot): benchmarking version of `map`. Input data
  needs dimensions (longitude, latitude).
- `benchmarking_timeseries` (1D plot): benchmarking version of
  `timeseries`. Input data needs single dimension time.
- `benchmarking_zonal` (2D plot): benchmarking version of
  `zonal_mean_profile`. Input data needs dimensions (latitude, altitude) or
  latitude, air_pressure.
- `diurnal_cycle` (1D plot): plot diurnal cycle. Input data needs single
  dimension hour.
- `hovmoeller_anncyc_vs_lat_or_lon` (2D plot): plot Hovmoeller plot. Input
  data needs dimensions (month_number, latitude) or (month_number,
  longitude).
- `hovmoeller_time_vs_lat_or_lon` (2D plot): plot Hovmoeller plot. Input data
  needs dimensions (time, latitude) or (time, longitude).
- `hovmoeller_z_vs_time` (2D plot): plot Hovmoeller plot. Input data needs
  dimensions (time, altitude) or (time, air_pressure).
- `map` (2D plot): plot map plot. Input data needs dimensions (longitude,
  latitude).
- `timeseries` (1D plot): plot time series. Input data needs single dimension
  time.
- `variable_vs_lat` (1D plot): plot variable vs. latitude. Input data needs
  single dimension latitude.
- `zonal_mean_profile` (2D plot): plot zonal mean profile. Input data needs
  dimensions (latitude, altitude) or latitude, air_pressure.

Each plot type belongs to one or more classes which define how the data is
plotted:

- Class *1D plot*: for each variable separately, all datasets are plotted in
  one single figure. Input data needs to be 1D.
- Class *2D plot*: for each variable and dataset, an individual figure is
  plotted. Input data needs to be 2D. A single reference dataset can be defined
  by setting the facet `reference_for_monitor_diags: True` in the dataset
  definition in the recipe. In this case, three panels are plotted, incl. a
  bias. Note that if a reference dataset is defined, all input datasets need to
  be given on the same horizontal and vertical grid (you can use the
  preprocessors [`esmvalcore.preprocessor.regrid()`](https://docs.esmvaltool.org/projects/ESMValCore/en/latest/api/esmvalcore.preprocessor.html#esmvalcore.preprocessor.regrid) and
  [`esmvalcore.preprocessor.extract_levels()`](https://docs.esmvaltool.org/projects/ESMValCore/en/latest/api/esmvalcore.preprocessor.html#esmvalcore.preprocessor.extract_levels) for this).
- Class *boxplot*: one figure for all datasets is plotted. Input data needs to
  be 0D.

In addition, for many plot types, a benchmarking version exists
(`benchmarking_*`). These plots include envelopes (1D plots) or hatching (2D
plots) to visualize data ranges (e.g., multi-model percentiles) in addition to
the data itself. For these plots, at least one benchmarking dataset needs to be
defined by setting the facet `benchmark_dataset: True` in the dataset
definition in the recipe. Reference datasets are not supported for 2D
benchmarking diagnostics.

## Author

Manuel Schlund (DLR, Germany)

<a id="diag-monitor-multi-datasets-recipe-options"></a>

## Configuration options in recipe

facet_used_for_labels: str, optional (default: ‘dataset’)
: Facet used to label different datasets in plot titles and legends. For
  example, `facet_used_for_labels: 'dataset'` will use dataset names in
  plot titles and legends; `facet_used_for_labels: 'exp'` will use
  experiments in plot titles and legends. In addition,
  `facet_used_for_labels` is used to select the correct `plot_kwargs` for
  the different datasets (see configuration options for the different plot
  types below).

figure_kwargs: dict, optional
: Optional keyword arguments for [`matplotlib.pyplot.figure()`](https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.figure.html#matplotlib.pyplot.figure). By
  default, uses `{constrained_layout: True}`.

group_variables_by: str, optional (default: ‘short_name’)
: Facet or coordinate which is used to create variable groups. For each
  variable group, an individual plot is created. Specifying a coordinate
  allows to create one plot for each point along a dimension. For example,
  when used in combination with the preprocessor function
  [`esmvalcore.preprocessor.extract_shape()`](https://docs.esmvaltool.org/projects/ESMValCore/en/latest/api/esmvalcore.preprocessor.html#esmvalcore.preprocessor.extract_shape) the shape_id coordinate
  can be used to create one plot for each shape.

matplotlib_rc_params: dict, optional
: Optional [`matplotlib.RcParams`](https://matplotlib.org/stable/api/matplotlib_configuration_api.html#matplotlib.RcParams) used to customize matplotlib plots.
  Options given here will be passed to [`matplotlib.rc_context()`](https://matplotlib.org/stable/api/matplotlib_configuration_api.html#matplotlib.rc_context) and used
  for all plots produced with this diagnostic. Note: fontsizes specified here
  might be overwritten by the plot-type-specific option `fontsize` (see
  below).

plots: dict
: Plot types plotted by this diagnostic (see list above). Dictionary keys
  must be elements of the list above.  Dictionary values are dictionaries
  used as options for the corresponding plot. The allowed options for the
  different plot types are given below.

plot_filename: str, optional
: Filename pattern for the plots. By default, uses
  `'{plot_type}_{real_name}_{dataset}_{mip}_{exp}_{ensemble}'`.  All tags
  (i.e., the entries in curly brackets, e.g., `'{dataset}'`, are replaced
  with the corresponding tags).

plot_folder: str, optional
: Path to the folder to store figures. By default, uses
  `'{plot_dir}/../../{dataset}/{exp}/{modeling_realm}/{real_name}'`.  All
  tags (i.e., the entries in curly brackets, e.g., `'{dataset}'`, are
  replaced with the corresponding tags). `'{plot_dir}'` is replaced with
  the default ESMValTool plot directory (i.e.,
  `output_dir/plots/diagnostic_name/script_name/`, see
  [Output](https://docs.esmvaltool.org/projects/ESMValCore/en/latest/quickstart/output.html#outputdata)).

savefig_kwargs: dict, optional
: Optional keyword arguments for [`matplotlib.pyplot.savefig()`](https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.savefig.html#matplotlib.pyplot.savefig). By
  default, uses `{bbox_inches: 'tight', dpi: 300, orientation:
  'landscape'}`.

seaborn_settings: dict, optional
: Options for [`seaborn.set_theme()`](https://seaborn.pydata.org/generated/seaborn.set_theme.html#seaborn.set_theme) (affects all plots). By default, uses
  `{style: 'ticks'}`.

<a id="diag-monitor-multi-datasets-plot-options"></a>

## Configuration options for 1D plots

aspect_ratio: float, optional (default: None)
: Aspect ratio of the plot.

axes_kwargs: dict, optional
: Optional calls to methods of the corresponding
  [`matplotlib.axes.Axes`](https://matplotlib.org/stable/api/_as_gen/matplotlib.axes.Axes.html#matplotlib.axes.Axes) instance. Dictionary keys are functions of
  [`matplotlib.axes.Axes`](https://matplotlib.org/stable/api/_as_gen/matplotlib.axes.Axes.html#matplotlib.axes.Axes). Dictionary values are used as argument(s)
  for these functions (if values are dictionaries, these are interpreted as
  keyword arguments; otherwise a single argument is assumed). String
  arguments can include facets in curly brackets which will be derived from
  the corresponding dataset, e.g., `{project}`, `{short_name}`,
  `{exp}`. Examples: `{set_title: 'Awesome Plot of {long_name}'}`,
  `{set_xlabel: '{short_name}'}`, `{set_xlim: [0, 5]}`.

caption: str, optional
: Figure caption used for provenance tracking. Can include facets in curly
  brackets which will be derived from the corresponding dataset, e.g.,
  `{project}`, `{short_name}`, `{exp}`. By default, uses a very basic
  caption.

envelope_kwargs: dict, optional
: Optional keyword arguments for [`iris.plot.fill_between()`](https://scitools-iris.readthedocs.io/en/latest/generated/api/iris.plot.html#iris.plot.fill_between). By default,
  uses `{alpha: 0.8, facecolor: 'lightblue', linewidth: 0.0, zorder: 1.0}`.
  Only relevant for benchmarking plots.

gridline_kwargs: dict, optional
: Optional keyword arguments for grid lines. By default, uses `{color:
  'lightgrey', alpha: 0.5}`. Use `gridline_kwargs: False` to not show grid
  lines.

hlines: list of dict, optional
: Horizontal lines to show in plot. Each list element corresponds to one
  line, and each list element should contain a dictionary with keywords
  arguments passed to [`matplotlib.axes.Axes.axhline()`](https://matplotlib.org/stable/api/_as_gen/matplotlib.axes.Axes.axhline.html#matplotlib.axes.Axes.axhline). Example: `[{y:
  0}, {y: 1, color: 'red'}]`.

legend_kwargs: dict, optional
: Optional keyword arguments for [`matplotlib.pyplot.legend()`](https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.legend.html#matplotlib.pyplot.legend). Use
  `legend_kwargs: False` to not show legends.

log_x: bool, optional (default: False)
: Use logarithmic X-axis.

log_y: bool, optional (default: False)
: Use logarithmic Y-axis.

plot_kwargs: dict, optional
: Optional keyword arguments for [`iris.plot.plot()`](https://scitools-iris.readthedocs.io/en/latest/generated/api/iris.plot.html#iris.plot.plot). Dictionary keys are
  elements identified by `facet_used_for_labels` or `'default'`, e.g.,
  `'CMIP6'` if `facet_used_for_labels: 'project'` or `'historical'` if
  `facet_used_for_labels: 'exp'`. Dictionary values are dictionaries used
  as keyword arguments for [`iris.plot.plot()`](https://scitools-iris.readthedocs.io/en/latest/generated/api/iris.plot.html#iris.plot.plot). String arguments can
  include facets in curly brackets which will be derived from the
  corresponding dataset, e.g., `{project}`, `{short_name}`, `{exp}`.
  Examples: `{default: {linestyle: '-', label: '{project}'}, CMIP6: {color:
  'red', linestyle: '--'}, OBS: {color: 'black'}}`.

pyplot_kwargs: dict, optional
: Optional calls to functions of [`matplotlib.pyplot`](https://matplotlib.org/stable/api/pyplot_summary.html#module-matplotlib.pyplot). Dictionary keys
  are functions of [`matplotlib.pyplot`](https://matplotlib.org/stable/api/pyplot_summary.html#module-matplotlib.pyplot). Dictionary values are used as
  argument(s) for these functions (if values are dictionaries, these are
  interpreted as keyword arguments; otherwise a single argument is assumed).
  String arguments can include facets in curly brackets which will be derived
  from the corresponding dataset, e.g., `{project}`, `{short_name}`,
  `{exp}`. Examples: `{title: 'Awesome Plot of {long_name}'}`, `{xlabel:
  '{short_name}'}`, `{xlim: [0, 5]}`.

rasterize: bool, optional (default: False)
: If `True`, use [rasterization](https://matplotlib.org/stable/gallery/misc/rasterization_demo.html) for plots to produce smaller files.  This
  is only relevant for vector graphics (e.g., `output_file_type: 'pdf'`).

time_format: str, optional (default: None)
: `strftime()` format string that is used to format
  the time axis using [`matplotlib.dates.DateFormatter`](https://matplotlib.org/stable/api/dates_api.html#matplotlib.dates.DateFormatter). If `None`,
  use the default formatting imposed by the iris plotting function.

transpose_axes: bool, optional (default: False)
: Swap X- and Y-axis.

x_major_formatter: str, optional (default: None)
: Format string for [`matplotlib.ticker.FormatStrFormatter`](https://matplotlib.org/stable/api/ticker_api.html#matplotlib.ticker.FormatStrFormatter) used to
  format major tick labels of X-axis.

x_minor_formatter: str, optional (default: None)
: Format string for [`matplotlib.ticker.FormatStrFormatter`](https://matplotlib.org/stable/api/ticker_api.html#matplotlib.ticker.FormatStrFormatter) used to
  format minor tick labels of X-axis.

y_major_formatter: str, optional (default: None)
: Format string for [`matplotlib.ticker.FormatStrFormatter`](https://matplotlib.org/stable/api/ticker_api.html#matplotlib.ticker.FormatStrFormatter) used to
  format major tick labels of Y-axis.

y_minor_formatter: str, optional (default: None)
: Format string for [`matplotlib.ticker.FormatStrFormatter`](https://matplotlib.org/stable/api/ticker_api.html#matplotlib.ticker.FormatStrFormatter) used to
  format minor tick labels of Y-axis.

## Configuration options for 2D plots

aspect_ratio: float, optional (default: None)
: Aspect ratio of the plot.

axes_kwargs: dict, optional
: Optional calls to methods of the corresponding
  [`matplotlib.axes.Axes`](https://matplotlib.org/stable/api/_as_gen/matplotlib.axes.Axes.html#matplotlib.axes.Axes) instance. Dictionary keys are functions of
  [`matplotlib.axes.Axes`](https://matplotlib.org/stable/api/_as_gen/matplotlib.axes.Axes.html#matplotlib.axes.Axes). Dictionary values are used as argument(s)
  for these functions (if values are dictionaries, these are interpreted as
  keyword arguments; otherwise a single argument is assumed). String
  arguments can include facets in curly brackets which will be derived from
  the corresponding dataset, e.g., `{project}`, `{short_name}`,
  `{exp}`. Examples: `{set_title: 'Awesome Plot of {long_name}'}`,
  `{set_xlabel: '{short_name}'}`, `{set_xlim: [0, 5]}`.

caption: str, optional
: Figure caption used for provenance tracking. Can include facets in curly
  brackets which will be derived from the corresponding dataset, e.g.,
  `{project}`, `{short_name}`, `{exp}`. By default, uses a very basic
  caption.

cbar_label: str, optional (default: ‘{short_name} [{units}]’)
: Colorbar label. Can include facets in curly brackets which will be derived
  from the corresponding dataset, e.g., `{project}`, `{short_name}`,
  `{exp}`.

cbar_label_bias: str, optional (default: ‘Δ{short_name} [{units}]’)
: Colorbar label for plotting biases. Can include facets in curly brackets
  which will be derived from the corresponding dataset, e.g., `{project}`,
  `{short_name}`, `{exp}`. Only relevant for plots including reference
  datasets.

cbar_kwargs: dict, optional
: Optional keyword arguments for [`matplotlib.pyplot.colorbar()`](https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.colorbar.html#matplotlib.pyplot.colorbar). By
  default, uses `{orientation: 'vertical'}`.

cbar_kwargs_bias: dict, optional
: Optional keyword arguments for [`matplotlib.pyplot.colorbar()`](https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.colorbar.html#matplotlib.pyplot.colorbar) for
  plotting biases. These keyword arguments update (and potentially overwrite)
  the `cbar_kwargs` for the bias plot. Only relevant for plots including
  reference datasets.

common_cbar: bool, optional (default: False)
: Use a common colorbar for the top panels (i.e., plots of the dataset and
  the corresponding reference dataset) when using a reference dataset. If
  neither `vmin` and `vmax` nor `levels` is given in `plot_kwargs`,
  the colorbar bounds are inferred from the dataset in the top left panel,
  which might lead to an inappropriate colorbar for the reference dataset
  (top right panel). Thus, specify `vmin` and `vmax` or `levels` when
  using `common_cbar: True`. Only relevant for plots including reference
  datasets.

fontsize: int, optional (default: None)
: Fontsize used for ticks, labels and titles. For the latter, use the given
  fontsize plus 2. Does not affect suptitles. If not given, use default
  matplotlib values. For a more fine-grained definition of fontsizes, use the
  option `matplotlib_rc_params` (see above).

gridline_kwargs: dict, optional (default: False)
: Optional keyword arguments for grid lines. Use `gridline_kwargs: False`
  to not show grid lines.

log_x: bool, optional (default: False)
: Use logarithmic X-axis.

log_y: bool, optional (default: False)
: Use logarithmic Y-axis.

plot_func: str, optional (default: ‘contourf’)
: Plot function used to plot the maps. Must be a function of [`iris.plot`](https://scitools-iris.readthedocs.io/en/latest/generated/api/iris.plot.html#module-iris.plot)
  that supports plotting of 2D data.

plot_kwargs: dict, optional
: Optional keyword arguments for the plot function defined by `plot_func`.
  Dictionary keys are elements identified by `facet_used_for_labels` or
  `'default'`, e.g., `'CMIP6'` if `facet_used_for_labels: 'project'` or
  `'historical'` if `facet_used_for_labels: 'exp'`. Dictionary values are
  dictionaries used as keyword arguments for the plot function defined by
  `plot_func`. String arguments can include facets in curly brackets which
  will be derived from the corresponding dataset, e.g., `{project}`,
  `{short_name}`, `{exp}`. Examples: `{default: {levels: 2}, CMIP6:
  {vmin: 200, vmax: 250}}`. In addition to the [normalization](https://matplotlib.org/stable/users/explain/colors/colormapnorms.html) options
  supported by the plot function, the option `{norm: 'centered'}` can be
  specified. In this case, the keywords `vcenter` and `halfrange` should
  be used instead of `vmin` or `vmax` (see
  [`CenteredNorm`](https://matplotlib.org/stable/api/_as_gen/matplotlib.colors.CenteredNorm.html#matplotlib.colors.CenteredNorm)).

plot_kwargs_bias: dict, optional
: Optional keyword arguments for the plot function defined by `plot_func`
  for plotting biases. These keyword arguments update (and potentially
  overwrite) the `plot_kwargs` for the bias plot. By default, uses `{cmap:
  'bwr', norm: 'centered'}`. Only relevant for plots including reference
  datasets.

projection: str, optional (default: None)
: Projection used for the plot. Needs to be a valid projection class of
  [`cartopy.crs`](https://cartopy.readthedocs.io/latest/reference/crs.html#module-cartopy.crs). Keyword arguments can be specified using the option
  `projection_kwargs`. For map plots, `'Robinson'` is used as default.

projection_kwargs: dict, optional
: Optional keyword arguments for the projection given by `projection`. For
  map plots, the default keyword arguments `{central_longitude: 10}` are
  used.

pyplot_kwargs: dict, optional
: Optional calls to functions of [`matplotlib.pyplot`](https://matplotlib.org/stable/api/pyplot_summary.html#module-matplotlib.pyplot). Dictionary keys
  are functions of [`matplotlib.pyplot`](https://matplotlib.org/stable/api/pyplot_summary.html#module-matplotlib.pyplot). Dictionary values are used as
  argument(s) for these functions (if values are dictionaries, these are
  interpreted as keyword arguments; otherwise a single argument is assumed).
  String arguments can include facets in curly brackets which will be derived
  from the corresponding dataset, e.g., `{project}`, `{short_name}`,
  `{exp}`. Examples: `{title: 'Awesome Plot of {long_name}'}`, `{xlabel:
  '{short_name}'}`, `{xlim: [0, 5]}`.

rasterize: bool, optional (default: False)
: If `True`, use [rasterization](https://matplotlib.org/stable/gallery/misc/rasterization_demo.html) for plots to produce smaller files.  This
  is only relevant for vector graphics (e.g., `output_file_type: 'pdf'`).

show_stats: bool, optional (default: True)
: Show basic statistics on the plots.

time_format: str, optional (default: None)
: `strftime()` format string that is used to format
  the time axis using [`matplotlib.dates.DateFormatter`](https://matplotlib.org/stable/api/dates_api.html#matplotlib.dates.DateFormatter). If `None`,
  use the default formatting imposed by the iris plotting function.

transpose_axes: bool, optional (default: False)
: Swap X- and Y-axis.

x_major_formatter: str, optional (default: None)
: Format string for [`matplotlib.ticker.FormatStrFormatter`](https://matplotlib.org/stable/api/ticker_api.html#matplotlib.ticker.FormatStrFormatter) used to
  format major tick labels of X-axis.

x_minor_formatter: str, optional (default: None)
: Format string for [`matplotlib.ticker.FormatStrFormatter`](https://matplotlib.org/stable/api/ticker_api.html#matplotlib.ticker.FormatStrFormatter) used to
  format minor tick labels of X-axis.

x_pos_stats_avg: float, optional (default: 0.01)
: Text X-position of average (shown on the left) in Axes coordinates. Can be
  adjusted to avoid overlap with the figure. Only relevant if `show_stats:
  True`.

x_pos_stats_bias: float, optional (default: 0.7)
: Text X-position of bias statistics (shown on the right) in Axes
  coordinates. Can be adjusted to avoid overlap with the figure. Only
  relevant if `show_stats: True`.

y_major_formatter: str, optional (default: None)
: Format string for [`matplotlib.ticker.FormatStrFormatter`](https://matplotlib.org/stable/api/ticker_api.html#matplotlib.ticker.FormatStrFormatter) used to
  format major tick labels of Y-axis.

y_minor_formatter: str, optional (default: None)
: Format string for [`matplotlib.ticker.FormatStrFormatter`](https://matplotlib.org/stable/api/ticker_api.html#matplotlib.ticker.FormatStrFormatter) used to
  format minor tick labels of Y-axis.

## Configuration options for boxplots

axes_kwargs: dict, optional
: Optional calls to methods of the corresponding
  [`matplotlib.axes.Axes`](https://matplotlib.org/stable/api/_as_gen/matplotlib.axes.Axes.html#matplotlib.axes.Axes) instance. Dictionary keys are functions of
  [`matplotlib.axes.Axes`](https://matplotlib.org/stable/api/_as_gen/matplotlib.axes.Axes.html#matplotlib.axes.Axes). Dictionary values are used as argument(s)
  for these functions (if values are dictionaries, these are interpreted as
  keyword arguments; otherwise a single argument is assumed). String
  arguments can include facets in curly brackets which will be derived from
  the corresponding dataset, e.g., `{project}`, `{short_name}`,
  `{exp}`. Examples: `{set_title: 'Awesome Plot of {long_name}'}`,
  `{set_xlabel: '{short_name}'}`, `{set_xlim: [0, 5]}`.

caption: str, optional
: Figure caption used for provenance tracking. Can include facets in curly
  brackets which will be derived from the corresponding dataset, e.g.,
  `{project}`, `{short_name}`, `{exp}`. By default, uses a very basic
  caption.

fontsize: int, optional (default: None)
: Fontsize used for ticks, labels and titles. For the latter, use the given
  fontsize plus 2. Does not affect suptitles. If not given, use default
  matplotlib values. For a more fine-grained definition of fontsizes, use the
  option `matplotlib_rc_params` (see above).

plot_kwargs: dict, optional
: Optional keyword arguments for [`seaborn.boxplot()`](https://seaborn.pydata.org/generated/seaborn.boxplot.html#seaborn.boxplot).

pyplot_kwargs: dict, optional
: Optional calls to functions of [`matplotlib.pyplot`](https://matplotlib.org/stable/api/pyplot_summary.html#module-matplotlib.pyplot). Dictionary keys
  are functions of [`matplotlib.pyplot`](https://matplotlib.org/stable/api/pyplot_summary.html#module-matplotlib.pyplot). Dictionary values are used as
  argument(s) for these functions (if values are dictionaries, these are
  interpreted as keyword arguments; otherwise a single argument is assumed).
  String arguments can include facets in curly brackets which will be derived
  from the corresponding dataset, e.g., `{project}`, `{short_name}`,
  `{exp}`. Examples: `{title: 'Awesome Plot of {long_name}'}`, `{xlabel:
  '{short_name}'}`, `{xlim: [0, 5]}`.

scatter_kwargs: dict, optional
: Optional keyword arguments for [`matplotlib.pyplot.scatter()`](https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.scatter.html#matplotlib.pyplot.scatter) (used to
  plot marks for benchmarking dataset). By default, uses `{marker: 'x', s:
  200, linewidths: 2, color: 'red', zorder: 3}`.

var_order: list of str, optional (default: None)
: Order in which variables are plotted. Must be elements identified via
  `group_variables_by`. If `None`, use order in which data is given.
