v5 redundant choice on right click parameter

[RichardHeenan]

If right click parameter there is a redundant "Constrain parameter to its curretn value", which if you use it then becomes quite difficult to remove, in any case it is not needed as you simply turn the parameter off.

Likewise if you highlight one parameter and ctrl/click to highlight a second one, the right click menu, normally used for setting up a mutual constraint has "Constrain parameters to their current values", which is also redundant.

I'm sure these were part of a bigger ticket somewhere, but have got overlooked (or crept back in).

[rozyczko]

it is not needed as you simply turn the parameter off.

I thought the main point of constraining a parameter to current value was to allow it in the fitting and see the magnitude of its error?

[RichardHeenan]

Summary of discussion at online Zoom meeting today -

There may perhaps be some merit to seeing what the error on the parameter would be if it were adjusting, noting of course that the error estimates on all the other adusting parameters should also increase. This might be of more use if the correlation or covariance matrix was routinely available for inspection.
See #994
and #571

Our decision was to leave the code as is for now, awaiting the new calculation of errors on constrained parameters.
See #1682

What would be more useful, but would require a lot of work, is to be able to enter the expected value of a parameter with an error, as a piece of experimental data, e.g. I know that say vol_frac = 0.4 +-0.02 so that piece of data is included in the least squares fit. However this is difficult to get right, so it influences the fit appropriately, in a similar way to the issues of say trying to fit SAXS and SANS simultaneously. (Part of the issue may be that the SAXS and SANS data points actually need correlated errors between adjacent points.) I first came across this as "Method of predicate observations" in gas electron diffraction - a wide angle diffraction method for molecular structures but no doubt the same thought has occured on other areas. [I was a postdoc in the Bartell group at Univ Michigan from Nov 1979 to Sept 1982.]

See this paper: L.S.Bartell, DJ.Romenesko & T.C.Wong, Chapter 4 in Molecular Structure by Diffraction Methods: Volume 3, edited by G A Sim, L E Sutton, Royal Socitey of Chemistry, 1975), which Google has fortunately scanned for us:.
https://books.google.co.uk/books?id=iHAoDwAAQBAJ&pg=PA72&lpg=PA72&dq=bartell+romenesko+wong&source=bl&ots=ZOc80DK-7h&sig=ACfU3U3rqwLy-Rb9scluyAi_vmy33NUC7w&hl=en&sa=X&ved=2ahUKEwjAyNXtr6juAhXIPsAKHYLdDhMQ6AEwDnoECBEQAg#v=onepage&q=bartell%20romenesko%20wong&f=false

The current sasview situation seems to be inconsistent depending what you have alreadty been doing. Sometimes using the "constrain to current value" gives an error estimate (perhaps only if that parameter has previously been adjusted, the value can seem spuriously small), sometimes not. The method is setting the min and max parameter values to a small range. Having chosen this option, it is sometimes hard to undo it, as sometimes right click on the parameter no longer brings up a menu! (Work around is to create a normal constraint in the constraints tab, then delete that.)
Perhaps some link here to #1063

[gonzalezma]

Most of Richard comments are still valid, but as we have decided to keep the option to constraint a parameter to its current value and there is not a clear way ahead to implement his suggestion "to be able to enter the expected value of a parameter with an error, as a piece of experimental data", should we close this?

[krzywon]

Was this addressed in either #2711 or #2743?