You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
SeriesRenderer to show both 1-sd error bar with DELTA_PLUS and 100% confidence interval with BIN_PLUS #160
Chris and Jeff would like to see both the 1-sd confidence (68%) level and 100% confidence levels on a plot. When both are found, the 100% confidence will always be a shaded region and the 68% confidence will be marked with error bars.
Actually the comment is backwards from what I was hoping for. The 1-sd confidence would be the shaded region, and the excursions out to bin min/max would be just the bars.
Some people might want the opposite since 1-sd is typically displayed by error bars, but since this is for engineering displays, I'm thinking more of a storage Oscilloscope display where if the time width is long compared to the screen resolution, the main variability of the signal is just blurred together, but the peaks stand out.
The idea is to try and capture what a display would look like with millions of points, without sending millions of points to the client.
The problem is when the min/max become dense, they occlude the stddev, which will always be smaller. One option would be to just do min max as filled region behind stddev as filled region (and potentially just skip the averages, since they will be at the center of the stddev).
Activity
rfe #160: prototype
cpiker commentedon Jan 25, 2026
Actually the comment is backwards from what I was hoping for. The 1-sd confidence would be the shaded region, and the excursions out to bin min/max would be just the bars.
Some people might want the opposite since 1-sd is typically displayed by error bars, but since this is for engineering displays, I'm thinking more of a storage Oscilloscope display where if the time width is long compared to the screen resolution, the main variability of the signal is just blurred together, but the peaks stand out.
The idea is to try and capture what a display would look like with millions of points, without sending millions of points to the client.
jbfaden commentedon Jan 26, 2026
This seems anologous to a box-and-whisker display. I'll play with this later today. It'll be a kludge, but there's a kludge in there either way.
jbfaden commentedon Jan 26, 2026
So if you don't mind the min-max bar being on top of the 1-stddev shaded region, it's a trivial change from what I had:
Actually I think that's pretty easy to fix, too. I'm not sure why I was thinking I'd have to visit the code twice.
VoyagerPWS commentedon Jan 26, 2026
The problem is when the min/max become dense, they occlude the stddev, which will always be smaller. One option would be to just do min max as filled region behind stddev as filled region (and potentially just skip the averages, since they will be at the center of the stddev).
cpiker commentedon Jan 26, 2026
So kind of like the peaks-n-averages display for Voyager, but now the average is replaced by (average +/- std_dev). Makes sense.
Removing the average line altogether makes sense. Zoom in enough and std. dev. goes to zero any thus recovering the original data points naturally.
jbfaden commentedon Jan 27, 2026
Here's a jar file with the change, if you are interested in playing with this today:
https://cottagesystems.com/jenkins/job/autoplot-release/lastSuccessfulBuild/artifact/autoplot/Autoplot/dist/autoplot.jar