Plot uncertainty and intervals

std.plot draws caller-supplied lower and upper bounds:

• geom_errorbar draws a vertical interval with end caps.
• geom_linerange draws the interval without caps.
• geom_ribbon fills the area between bounds, sorted by x within each group.

Map bounds with aes(x="time", ymin="lower", ymax="upper"). A y mapping is optional and ignored by these geoms, so they can inherit a centre-value mapping shared with points or a line. All three accept layer data/mapping overrides, colour, alpha, linewidth and na_rm. Ribbons also accept fill. They do not compute statistical estimates, errors or confidence levels.

Bounds and transformations

Both bounds must be continuous columns. Finite ymin > ymax is an error at construction. Equal bounds are valid. Both endpoints contribute to axis training, including in facets.

Position transforms and limits apply to each bound. A row with a missing, non-finite, out-of-domain or scale-excluded endpoint is omitted, with a note unless na_rm=True. Use coordinate limits to clip a visible interval without removing its data. Reversed axes and coord_flip preserve the interval.

Linewidth is in millimetres. Error-bar width is the total cap width in transformed x units (category spacing for discrete x); it defaults to 0.2 and currently must be in (0, 1]. On a log x axis it therefore describes a distance in log space. To display horizontal intervals, use coord_flip. Independent xmin/xmax mappings are not implemented.

Ribbon grouping and gaps

Ribbons require continuous x. Discrete colour/fill mappings or an explicit group define separate bands. Colour and fill must be constant within each continuous segment of a group; differing styles produce a rendering error instead of arbitrarily choosing one. Duplicate x values within a segment also produce an error. Use distinct x values for a well-defined band.

Missing bounds at a finite x split a ribbon into separate polygons. A segment needs at least two valid rows to form a polygon. Coordinates are joined by straight edges; no smoothing, interpolation between groups, or statistical fitting occurs. Ribbons use identity positioning, not stacking.