Histogram

Freedman-Diaconis is the default because it is robust to outliers. Sturges, Scott, Rice, square root, Doane and auto are available as named rules, and they reproduce the standard reference implementation exactly, so the chart can be stated in a methods section. A fixed number of bins or a fixed width covers the case where the audience expects a particular grid.

Round bin edges are on by default: the rule's width is snapped to 1, 2, 2.5 or 5 times a power of ten and the edges align to multiples of it, so bins read as 10 to 12 rather than 9.87 to 11.93. Bins are half-open with the last one closed, and a methods footnote names the rule that was used.

Yes, free. Type the edges you need, for example 0, 10, 12, 15, 20, and bins may differ in width; observations outside the first or last edge are not drawn and the footnote counts them. Or keep a fixed width and set where a bin starts, such as 0.5 for bins from 9.5 to 10.5.

Bars can show count, percent of group or density. A cumulative percentage line, a density curve and a fitted normal curve can be drawn over them; with unequal bin widths the curves appear only when the bars show density, because a count bar and a curve are not comparable then.

Reference lines for the mean, the median and up to four percentiles can be drawn for all data or per group. Percentiles follow the quartile definition you pick, with Excel / DAX inclusive as the default, so a P95 line on the chart is the same number PERCENTILE.INC returns in your model. Excel / DAX exclusive, Tukey's hinges and all nine Hyndman-Fan definitions are available.

Premium adds process capability. Type the spec limits for the whole chart, or map LSL, USL and Target measures from a spec table or product master: every group is then judged against its own specification, and the limits follow the slicers. The visual computes Cp, Cpk, Cpl, Cpu, Cpm, Pp and Ppk following the AIAG SPC manual, the share and ppm expected outside spec, and plus or minus 3 sigma control limits.

An index that cannot be computed is not shown, never guessed: too few observations, zero spread, crossed limits, a target outside the limits or an ambiguous limit give no index at all. Within-group sigma is taken in production order, so the Sample / ID field should sort the way the parts were made; the footnote names the field. Premium also adds significance between groups, weights, advanced density kernels and small multiples.