Opened 15 years ago
Closed 15 years ago
#46 closed task (fixed)
Interface for 2D fitting
Reported by: | srkline | Owned by: | srkline |
---|---|---|---|
Priority: | major | Milestone: | Analysis_4.0_Beta |
Component: | Analysis | Keywords: | |
Cc: | Blocking: | ||
Task: | DanseIntegration |
Description
Similar to the 2D interface for calculation, just more things to do since 2D fitting is a new topic. Some subtasks to tackle:
- FuncFit? operation specifics
- How to display fit
- how to display residuals
- ??
Change History (4)
comment:1 Changed 15 years ago by srkline
- Milestone set to Analysis_4.0_Beta
comment:2 Changed 15 years ago by srkline
comment:3 Changed 15 years ago by srkline
I'm having some real trouble with the fitting of the 2D data - it almost looks like an internal bug in Igor. Chi-squared is not being calculated correctly - so the fitting doesn't converge to the known solution for model data. Chi-squared is of the order 1e6, when I calculate it manually to be 1e-5! This reeks of an earlier problem where there were SP wave hidden in the calculation. I can't find any, so I'm a little bit stuck right now trying to duplicate the problem with a simpler system. Right now Cylinder2D is the only 2D function, and the only problem.
comment:4 Changed 15 years ago by srkline
- Resolution set to fixed
- Status changed from new to closed
Well, the fitting problem was not WM fault, but a combination of errors on my part where I was sloppy in the ordering of qx and qy in the parameter lists for the function as I specified, and in the XOP. If you call functions with what they expect, they work correctly...
The interface seems to be functional for now, and simply needs testing for polish as I see how users interact with the interface.
Closed for now. Tweaks will be opened as separarte tickets.
The first pass of the 2D fitting (one model function in the XOP) is ready for testing. The Curve fit Setup panel now has a checkbox for "2D Functions" that changes some of the button titles and corresponding procedures. Models are plotted as a 2D image, data and appended models are as a surface in Gizmo. Currently the fit and report are done as the Gizmo surface. The fitting works and behaves like the 1D fitting. visualization will need to be tweaked as we get user feedback.