Position along the main axis.
OptionalyAbsolute lower bound of the interval, when the chart draws one.
OptionalyAbsolute upper bound of the interval, when the chart draws one.
OptionalzName of the group this estimate belongs to, for a chart drawing an interval per group at each category.
Named to match LinePoint.z and carried for the same reason. A dodged error bar over two treatments puts two estimates at every category, and without this they arrive as two readings of one name with nothing telling them apart. The grouping cannot be recovered from emission order, because nothing states that order — so the comparison the chart exists to support, whether one group's interval overlaps another's, is the thing that goes missing (#942).
Meaningful on the grouped shape, ErrorBarPoint[][], where every point
in a series carries the same value. A single-series chart has one group
and needs no name for it.
The study's effect estimate.
Required here where ErrorBarPoint.y is optional. A forest plot is read by whether each interval crosses the null, and that question is only answerable relative to the estimate — a row without one is not a study with a missing number, it is not a forest plot row at all (#1047).
OptionalweightThe study's weight in the pooled estimate, as a fraction of one.
A forest plot encodes this as marker area, which is a magnitude a reader is otherwise never told: two studies whose intervals look alike can contribute wholly differently to the result.
OptionalpooledMarks the pooled summary rather than a study.
It is a different kind of row -- it is not evidence, it is what the evidence came to -- and announcing it as one more study invites a reader to count it among them.
One row of a forest plot: a study's effect estimate with its interval.
A meta-analysis draws one of these per study against a shared null line, with a pooled summary at the foot. It is an ErrorBarPoint laid out on a categorical row axis, plus the two things that make the figure a forest plot rather than a row of intervals.