I have noticed some oddities here too but mostly have ignored them and adjusted my plots to compensate. I mostly use my own variant "plotHistoMulti" - here: https://code.jsoftware.com/wiki/User:Devon_McCormick/myStats - but have not yet nailed down exactly what is off with the underlying.
On Sat, Apr 10, 2021 at 12:13 AM Gilles Kirouac <[email protected]> wrote: > > About the histogram verb in stats/base > > NB. The result is a list of counts of the number of data points in each > interval. > > Intervals are specified by the left argument. histogram uses dyadic I. > > The I. primitive defines (1+#x) intervals. See its clear definition > in Nuvoc. In short, it reports the index of the interval of the > corresponding data value. > > The primitive implies both __ and _ limits: > > _5 0 5 I. _2 _7 0 3 9 > 1 0 1 2 3 > > But histogram does not always correctly report on the last interval: > > _5 0 5 histogram _2 _7 0 3 9 > 1 2 1 0 > > Where has the last y value been classified? Five values were passed, > but only four counted. > > _5 0 5 _ histogram _2 _7 0 3 9 NB. OK > 1 2 1 1 > > As currently defined, the interval defined by {:x must include the > maximum value of y. > > Apparently the current definition comes from the (old) Vocabulary in > an example of dyadic I. But the definition there was appropriate for the > given example (or vice-versa). > > In a library, should we not have a more easily reliable tool? and > with better symmetry. > > ~ Gilles > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > -- Devon McCormick, CFA Quantitative Consultant ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
