https://bugs.kde.org/show_bug.cgi?id=501831
--- Comment #8 from realkpa...@gmail.com --- (In reply to Alexander Semke from comment #7) > (In reply to realkpavel from comment #5) > > (In reply to Alexander Semke from comment #4) > > > (In reply to realkpavel from comment #2) > > > > Created attachment 179661 [details] > > > > File where the histogram problem occurs > > > > > > > > Hello, > > > > attaching the file where I noticed this. Hope this helps. If I can help > > > > any > > > > other way, please let me know. > > > > > > Thank you! > > > > > > Which data did you use to reproduce this problem? I'm looking now at > > > Rsheet > > > from the LongStructures/Resistance spreadsheet the maximum is at 30.6849 > > > here. To what value did you increase the maximum to see this problem? > > > > All three histograms show this. In the one labeled "Resistance", the > > transition is at the high end (28.972 -> 28.973), In Rsheet it is 30.6849 -> > > 30.6850, in Rcontact it is 1.08622 -> 1.08623. In addition, changing the > > bin limits of Rcontact LongStructures/Worksheet breaks the binning change > > behavior of Rcontact (if you change the limits of Rsheet, then change the > > limits of Rcontact, the graph of Rcontact sometimes disappears) > > ok, I see it now. The behavior is correct since the convention used in GSL > and also in LabPlot is to define the last bin with strict < and not with <=. > > From GSL's documentation in > https://www.gnu.org/software/gsl/doc/html/histogram.html: > > "Thus any samples which fall on the upper end of the histogram are excluded. > If you want to include this value for the last bin you will need to add an > extra bin to your histogram." > > We'll add more tooltip texts in LabPlot in this area to explain this > behavior and to also properly document it in the documentation on > https://docs.labplot.org/ > > Can you please check and confirm this behavior on your side, too? Yes, that is the behavior I am seeing. A tooltip would certainly help. I understand the convention, it makes the bin widths the same mathematically, but I would disaggre with it if I was designing the system (since the default behavior guarantees one sample will be hidden, which will skew the data a lot in case of low sample numbers like here). But that is of course not my choice to make. Thank you for the work you're putting into this. -- You are receiving this mail because: You are watching all bug changes.
