I’m having an issue with TVirtualFitter:GetFitter()->GetConfidenceIntervals(h). This works fine when h has a relatively small number of bins but once it has more than roughly 100 then I run into an issue where many of the bins start being set to 0. It alternates between bins that are set properly and bins that are 0 in a regular pattern, the exact pattern depending on the total number of bins. I don’t see this behavior in 5.34 but unfortunately I have no choice but to use 5.22. Does anybody remember seeing this problem? Is there a workaround for people who don’t have a choice in what root version they use? Thanks for any help.
After trying some more examples it seems that the only bins that aren’t returned as zero are the bins which contain the centers of the histogram that is being fit. I looked at the GetConfidenceInterval code from 5.22 and there’s nothing there that I can imagine would cause this behavior. Pretty confusing.