Box Plot
Box Plot and Violin Plot analyze the same distribution: one numeric Value, optionally conditioned on Group by. Box Plot displays quartiles, whiskers, and outliers.
Controls
- Value: choose one objective or numeric parameter. Objectives appear first,
followed by numeric parameters.
Objective:andParameter:distinguish their source, including when names match. In multi-objective studies, select the objective to analyze explicitly. - Group by:
Noneshows one box for the selected Value. Choose a categorical parameter or an Optuna parameter explicitly declared as integer or stepped numeric to show one box per group. Continuous parameters are not grouped or automatically binned. Repeated or integer-looking observations alone do not make a numeric parameter eligible. - Normalize [0,1]: use the finite minimum and maximum of the entire selected Value column before splitting into groups. Constant columns become zero. Groups are not normalized independently.
Journal and Optuna database inputs supply declared distribution information. Numeric parameters in DesignExplorer-format CSV have no such metadata: they are available as Value, not numeric Group by. Parsed categorical CSV parameters can still group the Value.
The selections are independent for each widget. An unavailable Value falls back
to the first objective, or the first numeric parameter if there is no objective.
An unavailable Group by falls back to None.
Reading the plot
The Y-axis is the selected Value; with grouping, the X-axis identifies the parameter and group labels. Numeric groups are ordered numerically, categorical groups deterministically by label. Labels slant when needed and follow zooming and panning.
Missing or non-finite Values are excluded. With grouping, rows without a valid
grouping value are also excluded; Value and group always come from the same
trial. A real empty-string category is shown as (empty string), not treated as
missing. Distinct typed categories with the same ordinary label are identified
unambiguously, for example number: 1 and string: "1".
A box needs at least one finite Value. Singleton and constant groups are shown; groups with no finite Values are not. For smooth density shapes, use Violin Plot, which requires at least two finite, non-identical Values per group.
CSV export
CSV contains one summary row per displayed box: value, group_by, scale,
group, group_identity, n, mean, min, q1, median, q3, max,
whisker_low, whisker_high, and n_outliers. Selected columns and group
identities are unambiguous, and statistics use the displayed raw or normalized
scale. No data is exported when no box is displayed. group is the displayed
label; group_identity retains the category's original type or numeric group
value, so groups remain distinguishable even when CSV safety escaping changes a
label.
- Theory: Box Plot