plots: adjustable num_samples_per_point #407
Conversation
pmldrmota
changed the title
Added QSpinBox to context menu so that user can get more accurate error bars when averaging.
set_visible(False) didn't work (on Windows, at least).
pmldrmota
force-pushed
the
averaging-1d-adjustable-number-of-samples
branch
from
3760f42 to
64f62fa
Compare
| #: Keeps track of the running average and the number of samples therein. | ||
| self.averages_by_coords = dict[tuple[float, float], tuple[float, int]]() | ||
| self.averages_by_coords = dict[tuple[float, float], tuple[float, float, int]]() |
There was a problem hiding this comment.
Can we make this a a frozen dataclass/… instead of a tuple? This is pretty opaque to the casual reader.
There was a problem hiding this comment.
Certainly! I just pushed my WIP, but it should be using much of the same code as the 1D plot eventually.
|
I would like to improve the user experience a bit here, in the sense of making it clear to the average physicist what is actually going on here. If I hadn't thought about this particular problem already, I wouldn't have a clue what "samples per point" actually means. Perhaps we can make the label "Assume[d] samples per point" instead, to make clear that there are no samples involved in the averaging process itself? We could also add a tooltip with a brief explanation, and should probably move the discussion in the docstring to some user-visible documentation (maybe just a new "the ndscan applet" page?). Separately, I wonder whether it would be worth introducing a way to include enough information about the result distributions to do averaging and similar operations correctly directly with each result channel (either as channel metadata, or some kind of "rich FloatChannel"). |
Added QSpinBox to context menu so that user can get more accurate error bars when averaging.