Multi-dataset fitting interface: a few bugs and suggestions.

[Roman Tolchenov]

From James Lord:

1. One “bug”: I tried fitting with only one spectrum in the list and it gave an exception.

2. Another “bug” – removing spectra from the top of the list (or perhaps the middle) doesn’t remove the corresponding fit curves from the “list” so you get spectrum 2 displayed along with spectrum 1’s fit curve (and spectrum 1’s old difference). I’ve not yet checked what happens to the list of individual parameters, though stepping through the plots doesn’t update the values in the fit function pane, but it all sorts itself out when you hit “Fit” again.

3. If I fit with lots of workspaces there’s a Group “out_Workspaces” with the fit results. If I then reduce the number of workspaces and re-fit with the same fit output name, the surplus fit workspaces from the earlier fit are sometimes evicted from the group and now appear at the top level (out_Workspace_63 for example).

4. I note that adding additional spectra/workspaces gives them the same initial values as the ones already there, which is good (and sometimes an easy way to initialise things when the individual parameters are all to be free: initialise and fit two spectra, then add the remaining ones and fit again)

5. For multiple runs, usually we’d first reduce the data for each run to asymmetry (as the Muon Interface does). Then put the resulting series of 1-spectrum workspaces into the Multi dataset fit. So a way to select multiple workspaces (by range?) might be useful. Alternatively if I give it a Workspace Group can it take all the workspaces from it? (The workspace selection in “Add workspace” only shows individual workspaces including group members and not the groups themselves)

6. If I try to use a function with Attributes, all sorts of things go wrong, including the attribute not being set at all as far as the function is concerned. But having hard coded the attribute value into my Python fit function instead, it seems to work well. (My function doesn’t provide derivatives, so the numerical derivatives must be working).

7. In the “Edit parameter values” pop up window where a parameter isn’t global, Tab steps from one box to another in left-right order so it’s necessary to press Tab twice in between values, if filling them all in. Return exits the dialog (instead of going to the next parameter as in the main window). Down arrow goes to the box below but doesn’t activate it ready for typing in a number.

8. Should it even be possible to select the left column with the fit parameter indices? Would this column be more useful if it showed workspace names and spectrum numbers as in the top right of the main window?

[Nick Draper]

Moved to R3.5 at the R3.4 code freeze

[Stuart Campbell]

This ticket has been transferred to github issue ​11929