Opened 15 years ago
Closed 14 years ago
#54 closed bug (fixed)
USANS sub-matrix - assign or recalculate?
Reported by: | srkline | Owned by: | ajj |
---|---|---|---|
Priority: | major | Milestone: | Analysis_4.0_Beta |
Component: | Analysis | Keywords: | |
Cc: | Blocking: | ||
Task: | Resolution |
Description
As a separate issue to the "bad" last few points, is it correct to simply pull out a submatrix from the original calculation, or should the matrix be recalculated? Are end effects properly accounted for?
Change History (6)
comment:1 Changed 15 years ago by srkline
comment:2 Changed 15 years ago by srkline
As discussed, it seems like the proper solution is to recalculate the weighting matrix since the remainder term would be "cut" from the full matrix to make the sub matrix. The sub matrix is a nicer way, but we'll implement a re-calculation until we can prove otherwise
comment:3 Changed 14 years ago by srkline
After testing, it is clear that we need to recalculate the matrix, with a new slope, in an interactive manner.
comment:4 Changed 14 years ago by srkline
Note that recalculating the resolution matrix must also happen during global fitting!
comment:5 Changed 14 years ago by srkline
put the cursor information (pt1,pt2) in the wave note for the resolution, along with the (old) slope so that the code can detect a change and force a recalculation (and a new slope first)
comment:6 Changed 14 years ago by srkline
- Resolution set to fixed
- Status changed from new to closed
I think it's done.
- a copy of the original matrix (full data set ) is stored as the data set is loaded
- the matrix is recalculated (interactive slope), cursor positions are stored in the wave note
- original (full) matrix is restored if the cursors are un-checked
- propagated changes to the global fit
an ineresting note is that we cannot simply create a data mask wave (at the minimal penalty of calculating the function for the entire data set) since the FuncFit? call "trims" out the masked points and sends only the requested data to the function...
This would have been a nicer solution for this and for the Global fitting, but no.