source: sasview/src/sas/sasgui/perspectives/fitting/media/fitting_help.rst @ b7ce5ad

magnetic_scatt
Last change on this file since b7ce5ad was b7ce5ad, checked in by smk78, 5 years ago

Added section on Fitting Integer Parameters. Closes #882

  • Property mode set to 100644
File size: 32.2 KB
RevLine 
[ec392464]1.. fitting_help.rst
2
3.. This is a port of the original SasView html help file to ReSTructured text
4.. by S King, ISIS, during SasView CodeCamp-III in Feb 2015.
5
[b64b87c]6Fitting
7=======
[ec392464]8
[3e1c9e5]9.. note:: If some code blocks are not readable, expand the documentation window
10
[a6f3613]11.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
12
13Preparing to fit data
14---------------------
15
16To fit some data you must first load some data, activate one or more data sets,
[b64b87c]17send those data sets to fitting, and select a model to fit to each data set.
[a6f3613]18
[f3377b8]19Instructions on how to load and activate data are in the section :ref:`Loading_data`.
[ec392464]20
[a6f3613]21SasView can fit data in one of three ways:
[ec392464]22
[a6f3613]23*  in *Single* fit mode - individual data sets are fitted independently one-by-one
[ec392464]24
[5295cf5]25*  in *Simultaneous* fit mode - multiple data sets are fitted simultaneously to
26   the *same* model with/without constrained parameters (this might be useful,
27   for example, if you have measured the same sample at different contrasts)
[ec392464]28
[5005ae0]29*  in *Batch* fit mode - multiple data sets are fitted sequentially to the
30   *same* model (this might be useful, for example, if you have performed
31   a kinetic or time-resolved experiment and have *lots* of data sets!)
[a6f3613]32
33.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
[ec392464]34
[a6f3613]35Selecting a model
36-----------------
[ec392464]37
[5295cf5]38The models in SasView are grouped into categories. By default these consist of:
[ec392464]39
[5295cf5]40*  *Cylinder* - cylindrical shapes (disc, right cylinder, cylinder with endcaps
41   etc)
42*  *Ellipsoid* - ellipsoidal shapes (oblate,prolate, core shell, etc)
43*  *Parellelepiped* - as the name implies
44*  *Sphere* - sheroidal shapes (sphere, core multishell, vesicle, etc)
45*  *Lamellae* - lamellar shapes (lamellar, core shell lamellar, stacked
46   lamellar, etc)
[5005ae0]47*  *Shape-Independent* - models describing structure in terms of density
48   correlation functions, fractals, peaks, power laws, etc
[5295cf5]49*  *Paracrystal* - semi ordered structures (bcc, fcc, etc)
[a6f3613]50*  *Structure Factor* - S(Q) models
[5295cf5]51*  *Plugin Models* - User-created (custom/non-library) Python models
[ec392464]52
[a6f3613]53Use the *Category* drop-down menu to chose a category of model, then select
[f3377b8]54a model from the drop-down menu beneath. A graph of the chosen model, calculated
55using default parameter values, will appear. The graph will update dynamically
56as the parameter values are changed.
57
58You can decide your own model categorizations using the :ref:`Category_Manager`.
[ce62e75]59
[a6f3613]60Once you have selected a model you can read its help documentation by clicking
61on the *Description* button to the right.
62
63Show 1D/2D
64^^^^^^^^^^
65
66Models are normally fitted to 1D (ie, I(Q) vs Q) data sets, but some models in
67SasView can also be fitted to 2D (ie, I(Qx,Qy) vs Qx vs Qy) data sets.
68
69*NB: Magnetic scattering can only be fitted in SasView in 2D.*
70
71To activate 2D fitting mode, click the *Show 2D* button on the *Fit Page*. To
72return to 1D fitting model, click the same button (which will now say *Show 1D*).
[ce62e75]73
[ec392464]74.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
75
[a6f3613]76.. _Category_Manager:
[a0637de]77
[a6f3613]78Category Manager
79----------------
80
[f3377b8]81To change the model categorizations, either choose *Category Manager* from the
82*View* option on the menubar, or click on the *Modify* button on the *Fit Page*.
[a6f3613]83
[6aad2e8]84.. image:: cat_fig0.png
[a6f3613]85
[5295cf5]86The categorization of all models except the user supplied Plugin Models can be
87reassigned, added to, and removed using *Category Manager*. Models can also be
88hidden from view in the drop-down menus.
[a6f3613]89
[6aad2e8]90.. image:: cat_fig1.png
[a6f3613]91
[f3377b8]92Changing category
93^^^^^^^^^^^^^^^^^
[a6f3613]94
[5295cf5]95To change category, highlight a model in the list by left-clicking on its entry
96and then click the *Modify* button. Use the *Change Category* panel that appears
97to make the required changes.
[a6f3613]98
[6aad2e8]99.. image:: cat_fig2.png
[a6f3613]100
[f3377b8]101To create a category for the selected model, click the *Add* button. In order
102to delete a category, select the category name and click the *Remove Selected*
103button. Then click *Done*.
[a6f3613]104
[f3377b8]105Showing/hiding models
106^^^^^^^^^^^^^^^^^^^^^
[a6f3613]107
[5295cf5]108Use the *Enable All / Disable All* buttons and the check boxes beside each model
109to select the models to show/hide. To apply the selection, click *Ok*. Otherwise
110click *Cancel*.
[a6f3613]111
[8570246]112*NB: It may be necessary to change to a different category and then back again*
[f3377b8]113*before any changes take effect.*
[a6f3613]114
115.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
116
[f3377b8]117Model Functions
118---------------
[a6f3613]119
[5295cf5]120For a complete list of all the library models available in SasView, see
121the `Model Documentation <../../../index.html>`_ .
[a6f3613]122
[f3377b8]123It is also possible to add your own models.
[a6f3613]124
[f3377b8]125.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
[a6f3613]126
[b5846a10]127.. _Adding_your_own_models:
128
[3e1c9e5]129Adding your own Models
[f3377b8]130----------------------
[a6f3613]131
[3e1c9e5]132There are essentially three ways to generate new fitting models for SasView:
133
[5295cf5]134*  Using the SasView :ref:`New_Plugin_Model` helper dialog (best for beginners
135   and/or relatively simple models)
136*  By copying/editing an existing model (this can include models generated by
[5005ae0]137   the New Plugin Model* dialog) in the :ref:`Python_shell` or
[5295cf5]138   :ref:`Advanced_Plugin_Editor` (suitable for all use cases)
139*  By writing a model from scratch outside of SasView (only recommended for code
140   monkeys!)
[3e1c9e5]141
142Please read the guidance on :ref:`Writing_a_Plugin` before proceeding.
[f3377b8]143
[3e1c9e5]144**To be found by SasView your model must reside in the *~\\.sasview\\plugin_models* folder.**
[f3377b8]145
[a6f3613]146.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
147
[3e1c9e5]148.. _Plugin_Model_Operations:
[f3377b8]149
[3e1c9e5]150Plugin Model Operations
151-----------------------
[a6f3613]152
[3e1c9e5]153From the *Fitting* option in the menu bar, select *Plugin Model Operations*
[a6f3613]154
[82b0b05e]155.. image:: edit_model_menu.png
[a6f3613]156
[3e1c9e5]157and then one of the sub-options
[f3377b8]158
[3e1c9e5]159*  *New Plugin Model* - to create a plugin model template with a helper dialog
160*  *Sum|Multi(p1,p2)* - to create a plugin model by summing/multiplying *existing models* in the model library
161*  *Advanced Plugin Editor* - to create/edit a plugin model in a Python shell
162*  *Delete Plugin Models* - to delete a plugin model
163*  *Load Plugin Models* - to (re-)load plugin models
[a6f3613]164
[3e1c9e5]165.. _New_Plugin_Model:
166
167New Plugin Model
168^^^^^^^^^^^^^^^^
[a6f3613]169
[5295cf5]170Relatively straightforward models can be programmed directly from the SasView
171GUI using the *New Plugin Model Function*.
[26c8be3]172
[6aad2e8]173.. image:: new_model.png
[a6f3613]174
[5005ae0]175When using this feature, be aware that even if your code has errors, including
176syntax errors, a model file is still generated. When you then correct the errors
177and click 'Apply' again to re-compile you will get an error informing you that
178the model already exists if the 'Overwrite' box is not checked. In this case you
179will need to supply a new model function name. By default the 'Overwrite' box is
[26c8be3]180*checked*\ .
[f3377b8]181
[26c8be3]182A model file generated by this option can be viewed and further modified using
183the :ref:`Advanced_Plugin_Editor` .
[05829fb]184
[8b89396]185Note that the New Plugin Model Feature currently does not allow for parameters
186to be polydisperse.  However they can be edited in the Advanced Editor.
187
188
189**SasView version 4.2** made it possible to specify whether a plugin created
190with the *New Plugin Model* dialog is actually a form factor P(Q) or a structure
191factor S(Q). To do this, simply add one or other of the following lines under
192the *import* statements.
[f485ba0]193
194For a form factor::
195
196     form_factor = True
[5005ae0]197
[f485ba0]198or for a structure factor::
199
200     structure_factor = True
[5005ae0]201
[8b89396]202If the plugin is a structure factor it is *also* necessary to add two variables
203to the parameter list::
[f485ba0]204
[5005ae0]205     parameters = [
[ca383a0]206                     ['radius_effective', '', 1, [0.0, numpy.inf], 'volume', ''],
207                     ['volfraction', '', 1, [0.0, 1.0], '', ''],
[e081946]208                     [...],
[f485ba0]209
210and to the declarations of the functions Iq and Iqxy:::
211
212     def Iq(x , radius_effective, volfraction, ...):
213
214     def Iqxy(x, y, radius_effective, volfraction, ...):
215
[5005ae0]216Such a plugin should then be available in the S(Q) drop-down box on a FitPage (once
[f485ba0]217a P(Q) model has been selected).
218
[a6f3613]219Sum|Multi(p1,p2)
220^^^^^^^^^^^^^^^^
221
[6aad2e8]222.. image:: sum_model.png
[a0637de]223
[5295cf5]224This option creates a custom Plugin Model of the form::
[f3377b8]225
[5295cf5]226     Plugin Model = scale_factor * {(scale_1 * model_1) +/- (scale_2 * model_2)} + background
[f3377b8]227
[3e1c9e5]228or::
229
[f485ba0]230     Plugin Model = scale_factor * (model1 * model2) + background
[3e1c9e5]231
232In the *Easy Sum/Multi Editor* give the new model a function name and brief
233description (to appear under the *Details* button on the *FitPage*). Then select
[f3377b8]234two existing models, as p1 and p2, and the required operator, '+' or '*' between
[f485ba0]235them. Finally, click the *Apply* button to generate and test the model and then click *Close*.
[f3377b8]236
[5005ae0]237Any changes to a plugin model generated in this way only become effective *after* it is re-selected
238from the plugin models drop-down menu on the FitPage. If the model is not listed you can force a
[f485ba0]239recompilation of the plugins by selecting *Fitting* > *Plugin Model Operations* > *Load Plugin Models*.
240
[5005ae0]241**SasView version 4.2** introduced a much simplified and more extensible structure for plugin models
242generated through the Easy Sum/Multi Editor. For example, the code for a combination of a sphere model
[f485ba0]243with a power law model now looks like this::
244
245     from sasmodels.core import load_model_info
246     from sasmodels.sasview_model import make_model_from_info
[5005ae0]247
[f485ba0]248     model_info = load_model_info('sphere+power_law')
[e081946]249     model_info.name = 'MyPluginModel'
[f485ba0]250     model_info.description = 'sphere + power_law'
251     Model = make_model_from_info(model_info)
252
[5005ae0]253To change the models or operators contributing to this plugin it is only necessary to edit the string
254in the brackets after *load_model_info*, though it would also be a good idea to update the model name
[f485ba0]255and description too!!!
256
[5005ae0]257The model specification string can handle multiple models and combinations of operators (+ or *) which
258are processed according to normal conventions. Thus 'model1+model2*model3' would be valid and would
259multiply model2 by model3 before adding model1. In this example, parameters in the *FitPage* would be
260prefixed A (for model2), B (for model3) and C (for model1). Whilst this might appear a little
261confusing, unless you were creating a plugin model from multiple instances of the same model the parameter
[e081946]262assignments ought to be obvious when you load the plugin.
[f485ba0]263
[5005ae0]264If you need to include another plugin model in the model specification string, just prefix the name of
[e081946]265that model with *custom*. For instance::
266
267     sphere+custom.MyPluginModel
268
269To create a P(Q)*\S(Q) model use the @ symbol instead of * like this::
270
271     sphere@hardsphere
[5005ae0]272
273This streamlined approach to building complex plugin models from existing library models, or models
274available on the *Model Marketplace*, also permits the creation of P(Q)*\S(Q) plugin models, something
275that was not possible in earlier versions of SasView.
[f3377b8]276
[3e1c9e5]277.. _Advanced_Plugin_Editor:
[f3377b8]278
[3e1c9e5]279Advanced Plugin Editor
280^^^^^^^^^^^^^^^^^^^^^^
[ec392464]281
[3e1c9e5]282Selecting this option shows all the plugin models in the plugin model folder, on Windows this is
[f3377b8]283
[afb93df]284  *C:\\Users\\{username}\\.sasview\\plugin_models*
[f3377b8]285
286You can edit, modify, and save the Python code in any of these models using the
[3e1c9e5]287*Advanced Plugin Model Editor*. Note that this is actually the same tool as the :ref:`Python_shell` .
[f3377b8]288
[afb93df]289For details of the SasView plugin model format see :ref:`Writing_a_Plugin` .
[05829fb]290
[afb93df]291.. note:: Model files generated with the Sum/Multi option are still using the SasView 3.x model format. Unless you are confident about what you are doing, it is recommended that you only modify lines denoted with the ## <----- comments!
[f3377b8]292
[3e1c9e5]293When editing is complete, select *Run* > *Check Model* from the *Advanced Plugin Model Editor* menu bar. An *Info* box will appear with the results of the compilation and model unit tests. The model will only be usable if the tests 'pass'.
[f3377b8]294
[3e1c9e5]295.. image:: ../calculator/new_pycrust_example_2.png
296
297To use the model, go to the relevant *Fit Page*, select the *Plugin Models*
[f3377b8]298category and then select the model from the drop-down menu.
299
[3e1c9e5]300Any changes to a plugin model generated in this way only become effective *after* it is re-selected from the model drop-down menu on the FitPage.
[a6f3613]301
[3e1c9e5]302Delete Plugin Models
303^^^^^^^^^^^^^^^^^^^^
[a6f3613]304
[3e1c9e5]305Simply highlight the plugin model to be removed. The operation is final!!!
[f3377b8]306
[5295cf5]307*NB: Models shipped with SasView cannot be removed in this way.*
[ec392464]308
[3e1c9e5]309Load Plugin Models
310^^^^^^^^^^^^^^^^^^
[afb93df]311
[5295cf5]312This option loads (or re-loads) all models present in the
313*~\\.sasview\\plugin_models* folder.
[afb93df]314
[ec392464]315.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
316
[4666660]317.. _Fitting_Options:
318
[f3377b8]319Fitting Options
320---------------
[a6f3613]321
[f3377b8]322It is possible to specify which optimiser SasView should use to fit the data, and
323to modify some of the configurational parameters for each optimiser.
324
325From *Fitting* in the menu bar select *Fit Options*, then select one of the following
326optimisers:
[a6f3613]327
[f3377b8]328*  DREAM
329*  Levenberg-Marquardt
330*  Quasi-Newton BFGS
331*  Differential Evolution
332*  Nelder-Mead Simplex
[a6f3613]333
[6af2785]334The DREAM optimiser is the most sophisticated, but may not necessarily be the best
335option for fitting simple models. If uncertain, try the Levenberg-Marquardt optimiser
336initially.
337
[f3377b8]338These optimisers form the *Bumps* package written by P Kienzle. For more information
[4666660]339on each optimiser, see the :ref:`Fitting_Documentation`.
[a6f3613]340
[f3377b8]341.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
[a6f3613]342
[b7ce5ad]343Fitting Integer Parameters
344--------------------------
345
346Most of the parameters in SasView models will naturally take floating point (decimal)
347values, but there are some parameters which can only have integer values. Examples
348include, but are not limited to, the number of shells in a multilayer vesicle, the
349number of beads in a pearl necklace, the number of arms of a star polymer, and so on.
350Wherever possible/recognised, the integer nature of a parameter is specified in the
351respective model documentation and/or parameter table, so read the documentation
352carefully!
353
354Integer parameters must be fitted with care.
355
356Start with your best possible guess for the value of the parameter. And using
357*a priori* knowledge, fix as many of the other parameters as possible.
358 
359The SasView optimizers treat integer parameters internally as floating point
360numbers, but the values presented to the user are truncated or rounded, as
361appropriate.
362
363In most instances integer parameters will probably be greater than zero. A good
364policy in such cases is to use a constraint to enforce this.
365
366Because an integer parameter should, by definition, only move in integer steps,
367problems may be encountered if the optimizer step size is too small. Similarly,
368be **very careful** about applying polydispersity to integer parameters.
369
370The Levenberg-Marquardt and Quasi-Newton BFGS (and other derivative-based)
371optimizers are probably best avoided for fitting models with integer parameters.
372
373.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
374
[5ac0a7a]375Fitting Limits
376--------------
377
378By default, *SasView* will attempt to model fit the full range of the data; ie,
379across all *Q* values. If necessary, however, it is possible to specify only a
380sub-region of the data for fitting.
381
382In a *FitPage* or *BatchPage* change the *Q* values in the *Min* and/or *Max*
383text boxes. Vertical coloured bars will appear on the graph with the data and
384'theory' indicating the current *Q* limits (red = *Qmin*, purple = *Qmax*).
385
386To return to including all data in the fit, click the *Reset* button.
387
388.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
389
390
[f3377b8]391Shortcuts
392---------
[a6f3613]393
[f3377b8]394Copy/Paste Parameters
395^^^^^^^^^^^^^^^^^^^^^
[a6f3613]396
[f3377b8]397It is possible to copy the parameters from one *Fit Page* and to paste them into
398another *Fit Page* using the same model.
[a6f3613]399
[f3377b8]400To *copy* parameters, either:
[a6f3613]401
[f3377b8]402*  Select *Edit -> Copy Params* from the menu bar, or
403*  Use Ctrl(Cmd on Mac) + Left Mouse Click on the *Fit Page*.
[a6f3613]404
[f3377b8]405To *paste* parameters, either:
406
407*  Select *Edit -> Paste Params* from the menu bar, or
408*  Use Ctrl(Cmd on Mac) + Shift + Left-click on the *Fit Page*.
409
410If either operation is successful a message will appear in the info line at the
411bottom of the SasView window.
412
413Bookmark
414^^^^^^^^
[a6f3613]415
[1fda506]416To *Bookmark* a *Fit Page* either:
[a6f3613]417
[f3377b8]418*  Select a *Fit Page* and then click on *Bookmark* in the tool bar, or
419*  Right-click and select the *Bookmark* in the popup menu.
[a6f3613]420
421.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
422
[4666660]423.. _Status_bar:
424
[1fda506]425Status Bar & Console
426--------------------
427
428The status bar is located at the bottom of the SasView window and displays
429messages, hints, warnings and errors.
430
431At the right-hand side of the status bar is a button marked *Console*. The *Console*
432displays available message history and some run-time traceback information.
433
434During a long task the *Console* can also be used to monitor the progress.
435
436.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
437
[6af2785]438.. _Single_Fit_Mode:
439
[a6f3613]440Single Fit Mode
441---------------
[ec392464]442
[4666660]443*NB: Before proceeding, ensure that the Single Mode radio button at the bottom of*
444*the Data Explorer is checked (see the section* :ref:`Loading_data` *).*
445
[63d314b]446This mode fits one data set.
447
[b64b87c]448When data is sent to the fitting it is plotted in a graph window as markers.
[4666660]449
450If a graph does not appear, or a graph window appears but is empty, then the data
451has not loaded correctly. Check to see if there is a message in the :ref:`Status_Bar`
452or in the *Console* window.
453
454Assuming the data has loaded correctly, when a model is selected a green model
455calculation (or what SasView calls a 'Theory') line will appear in the earlier graph
456window, and a second graph window will appear displaying the residuals (the
457difference between the experimental data and the theory) at the same X-data values.
[f9250dc]458See :ref:`Assessing_Fit_Quality`.
[4666660]459
[a7c6f38]460The objective of model-fitting is to find a *physically-plausible* model, and
461set of model parameters, that generate a theory that reproduces the experimental
462data and minimizes the values of the residuals.
[4666660]463
464Change the default values of the model parameters by hand until the theory line
[a7c6f38]465starts to represent the experimental data. Then check the tick boxes alongside
[84ac3f1]466the 'background' and 'scale' parameters. Click the *Fit* button. SasView
[a7c6f38]467will optimise the values of the 'background' and 'scale' and also display the
468corresponding uncertainties on the optimised values.
[4666660]469
[a7c6f38]470.. note::
[84ac3f1]471   If the uncertainty on a fitted parameter is unrealistically large, or if it
472   displays as NaN, the model is most likely a poor representation of the data,
473   the parameter in question is highly correlated with one or more of the other
474   fitted parameters, or the model is relatively insensitive to the value of
475   that particular parameter.
[4666660]476
[47ace50]477In the bottom left corner of the *Fit Page* is a box displaying a normalised
478value of the statistical $\chi^2$ parameter (the reduced $\chi^2$,
479See :ref:`Assessing_Fit_Quality`) returned by the optimiser.
[4666660]480
481Now check the box for another model parameter and click *Fit* again. Repeat this
[84ac3f1]482process until all relevant parameters are checked and have been optimised. As
483the fit of the theory to the experimental data improves, the value of 'Reduced
484Chi2' will decrease. A good model fit should produce values of Reduced Chi2
485close to one, and certainly << 100. See :ref:`Assessing_Fit_Quality`.
[4666660]486
487SasView has a number of different optimisers (see the section :ref:`Fitting_Options`).
488The DREAM optimiser is the most sophisticated, but may not necessarily be the best
489option for fitting simple models. If uncertain, try the Levenberg-Marquardt optimiser
490initially.
[ec392464]491
492.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
493
[a6f3613]494Simultaneous Fit Mode
495---------------------
[ec392464]496
[63d314b]497*NB: Before proceeding, ensure that the Single Mode radio button at the bottom of*
498*the Data Explorer is checked (see the section* :ref:`Loading_data` *).*
[ec392464]499
[84ac3f1]500This mode is an extension of the :ref:`Single_Fit_Mode` that allows for some
501relatively extensive constraints between fitted parameters in a single *FitPage*
502or between several *FitPage*'s (eg, to constrain all fitted parameters to be the
503same in a contrast series of *FitPages* except for the solvent sld parameter,
[47ace50]504constrain the length to be twice that of the radius in a single *FitPage*,
[84ac3f1]505fix the radius of the sphere in one *FitPage* to be the same as the radius of
506the cylinder in a second *FitPage*, etc).
[63d314b]507
[b5846a10]508If the data to be fit are in multiple files, load each file, then select each file
509in the *Data Explorer*, and *Send To Fitting*. If multiple data sets are in one file,
510load that file, *Unselect All Data*, select just those data sets to be fitted, and
511*Send To Fitting*. Either way, the result should be that for *n* data sets you have
[6af2785]5122\ *n* graphs (*n* of the data and model fit, and *n* of the resulting residuals). But
[f9250dc]513it may be helpful to minimise the residuals plots for clarity. Also see
514:ref:`Assessing_Fit_Quality`.
[b5846a10]515
[5295cf5]516*NB: If you need to use a custom Plugin Model, you must ensure that model is
517available first (see* :ref:`Adding_your_own_models` *).*
[b5846a10]518
[6af2785]519Method
520^^^^^^
521
[b5846a10]522Now go to each *FitPage* in turn and:
523
524  Select the required category and model;
525
526  Unselect all the model parameters;
527
528  Enter some starting guesses for the parameters;
529
530  Enter any parameter limits (recommended);
[63d314b]531
[b5846a10]532  Select which parameters will refine (selecting all is generally a bad idea...);
[63d314b]533
[b5846a10]534When done, select *Constrained or Simultaneous Fit* under *Fitting* in the menu bar.
535
536In the *Const & Simul Fit* page that appears, select which data sets are to be
537simultaneously fitted (this will probably be all of them or you would not have loaded
538them in the first place!).
539
540To tie parameters between the data sets with constraints, check the 'yes' radio button
541next to *Add Constraint?* in the *Fit Constraints* box.
542
[84ac3f1]543To constrain all identically named parameters to fit *simultaneously* to the
544same value across all the *Fitpages* use the *Easy Setup* drop-down buttons in
545the *Const & Simul Fit* page.
[ec392464]546
[84ac3f1]547*NB: You can only constrain parameters that are set to refine.*
[b5846a10]548
[6af2785]549Constraints will generally be of the form
550
551  Mi Parameter1 = Mj.Parameter1
552
553however the text box after the '=' sign can be used to adjust this
554relationship; for example
555
556  Mi Parameter1 = scalar \* Mj.Parameter1
557
558A 'free-form' constraint box is also provided.
[b5846a10]559
560Many constraints can be entered for a single fit.
561
[84ac3f1]562When ready, click the *Fit* button on the *Const & Simul Fit* page, NOT the *Fit*
563button on the individual *FitPage*'s.
564
[b5846a10]565The results of the model-fitting will be returned to each of the individual
566*FitPage*'s.
[ec392464]567
[a7c6f38]568Note that the Reduced Chi2 value returned is the SUM of the Reduced Chi2 of
569each fit. To see the Reduced Chi2 value for a specific *FitPage*, click the
570*Compute* button at the bottom of that *FitPage* to recalculate. Note that in
[84ac3f1]571doing so the degrees of freedom will be set to Npts.
[a7c6f38]572See :ref:`Assessing_Fit_Quality`.  Moreover in the case of constraints the
[84ac3f1]573degrees of freedom are less than one might think due to those constraints.
574
[ec392464]575.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
576
[e6c74e8]577.. _Batch_Fit_Mode:
578
[a6f3613]579Batch Fit Mode
580--------------
[ec392464]581
[6af2785]582*NB: Before proceeding, ensure that the Single Mode radio button at the bottom of*
583*the Data Explorer is checked (see the section* :ref:`Loading_data` *). The Batch*
584*Mode button will be used later on!*
[63d314b]585
[6af2785]586This mode *sequentially* fits two or more data sets *to the same model*. Unlike in
587simultaneous fitting, in batch fitting it is not possible to constrain fit parameters
588between data sets.
[ec392464]589
[6af2785]590If the data to be fit are in multiple files, load each file in the *Data Explorer*.
591If multiple data sets are in one file, load just that file. *Unselect All Data*, then
592select a single initial data set to be fitted. Fit that selected data set as described
[5295cf5]593above under :ref:`Single_Fit_Mode`.
[ec392464]594
[5295cf5]595*NB: If you need to use a custom Plugin Model, you must ensure that model is
596available first (see* :ref:`Adding_your_own_models` *).*
[ec392464]597
[6af2785]598Method
599^^^^^^
[ec392464]600
[6af2785]601Now *Select All Data* in the *Data Explorer*, check the *Batch Mode* radio button
602at the bottom of that panel and *Send To Fitting*. A *BatchPage* will be created.
[ec392464]603
[6aad2e8]604.. image:: batch_button_area.png
[ec392464]605
[6af2785]606*NB: The Batch Page can also be created by checking the Batch Mode radio button*
607*and selecting New Fit Page under Fitting in the menu bar.*
[ec392464]608
[6af2785]609Using the drop-down menus in the *BatchPage*, now set up the *same* data set
610with the *same* model that you just fitted in single fit mode. A quick way to
611set the model parameter values is to just copy them from the earlier Single
612Fit. To do this, go back to the Single Fit *FitPage*, select *Copy Params*
613under *Edit* in the menu bar, then go back to the *BatchPage* and *Paste Params*.
[ec392464]614
[6af2785]615When ready, use the *Fit* button on the *BatchPage* to perform the fitting, NOT
616the *Fit* button on the individual *FitPage*'s.
[ec392464]617
[6af2785]618Unlike in single fit mode, the results of batch fits are not returned to
619the *BatchPage*. Instead, a spreadsheet-like :ref:`Grid_Window` will appear.
[ec392464]620
[6af2785]621If you want to visually check a graph of a particular fit, click on the name of
622a *Data set* in the *Grid Window* and then click the *View Fits* button. The
623data and the model fit will be displayed. If you select mutliple data sets they
624will all appear on one graph.
[ec392464]625
[6aad2e8]626.. image:: view_button.png
[ec392464]627
[6af2785]628*NB: In theory, returning to the BatchPage and changing the name of the I(Q)*
629*data source should also work, but at the moment whilst this does change the*
630*data set displayed it always superimposes the 'theory' corresponding to the*
631*starting parameters.*
[ec392464]632
[6af2785]633If you select a 'Chi2' value and click the *View Fits* button a graph of the
634residuals for that data set is displayed. Again, if you select multiple 'Chi2'
[f9250dc]635values then all the residuals data will appear on one graph. Also see
636:ref:`Assessing_Fit_Quality`.
[ec392464]637
[6af2785]638Chain Fitting
639^^^^^^^^^^^^^
[ec392464]640
[6af2785]641By default, the *same* parameter values copied from the initial single fit into
642the *BatchPage* will be used as the starting parameters for all batch fits. It
643is, however, possible to get *SasView* to use the results of a fit to a preceding
644data set as the starting parameters for the next fit in the sequence. This
645variation of batch fitting is called *Chain Fitting*, and will considerably speed
646up model-fitting if you have lots of very similar data sets where a few parameters
647are gradually changing. Do not use chain fitting on disparate data sets.
[ec392464]648
[6af2785]649To use chain fitting, select *Chain Fitting* under *Fitting* in the menu bar. It
650toggles on/off, so selecting it again will switch back to normal batch fitting.
[ec392464]651
[6af2785]652.. _Grid_Window:
[ec392464]653
[6af2785]654Grid Window
655^^^^^^^^^^^
[ec392464]656
[6af2785]657The *Grid Window* provides an easy way to view the results from batch fitting.
658It will be displayed automatically when a batch fit completes, but may be
659opened at any time by selecting *Show Grid Window* under *View* in the menu
660bar.
[ec392464]661
[6aad2e8]662.. image:: restore_batch_window.png
[ec392464]663
[6af2785]664Once a batch fit is completed, all model parameters are displayed but *not*
665their uncertainties. To view the uncertainties, click on a given column then
666go to *Edit* in the menu bar, select *Insert Column Before* and choose the
667required data from the list. An empty column can be inserted in the same way.
[ec392464]668
[6af2785]669To remove a column from the grid, click on the column header and choose
670*Remove Column* under *Edit* in the menu bar. The same functionality also
671allows you to re-order columns.
[ec392464]672
[6af2785]673*NB: You cannot insert/remove/re-order the rows in the Grid Window.*
[ec392464]674
[6af2785]675All of the above functions are also available by right-clicking on a column
676label.
[ec392464]677
[6aad2e8]678.. image:: edit_menu.png
[ec392464]679
[6af2785]680*NB: If there is an existing Grid Window and another batch fit is performed,*
681*an additional 'Table' tab will be added to the Grid Window.*
[ec392464]682
[6af2785]683The parameter values in the *currently selected* table of the *Grid Window*
684can be output to a CSV file by choosing *Save As* under *File* in the (*Grid*
685*Window*) menu bar. The default filename includes the date and time that the
686batch fit was performed.
[ec392464]687
[6af2785]688Saved CSV files can be reloaded by choosing *Open* under *File* in the *Grid*
689*Window* menu bar. The loaded parameters will appear in a new table tab.
[ec392464]690
[6aad2e8]691.. image:: file_menu.png
[ec392464]692
[6af2785]693*NB: Saving the Grid Window does not save any experimental data, residuals*
694*or actual model fits. Consequently if you reload a saved CSV file the*
695*ability to View Fits will be lost.*
[ec392464]696
[6af2785]697Parameter Plots
698^^^^^^^^^^^^^^^
[ec392464]699
[6af2785]700Any column of *numeric* parameter values can be plotted against another using
701the *Grid Window*. Simply select one column at the time and click the *Add*
702button next to the required *X/Y-axis Selection Range* text box. When both
703the X and Y axis boxes have been completed, click the *Plot* button.
[ec392464]704
[6af2785]705When the *Add* button is clicked, *SasView* also automatically completes the
706*X/Y-axis Label* text box with the heading from Row 1 of the selected table,
707but different labels and units can be entered manually.
[ec392464]708
[6aad2e8]709.. image:: plot_button.png
[ec392464]710
[6af2785]711The *X/Y-axis Selection Range* can be edited manually. The text control box
712recognises the operators +, -, \*, /, or 'pow', and allows the following
713types of expression :
[5005ae0]714
[6af2785]715  1) if an axis label range is a function of 1 or more *columns*, write
716     this type of expression
[ec392464]717
[6af2785]718     constant1 * column_name1 [minimum row index :  maximum  row index]
719     operator constant2 * column_name2 [minimum row index :  maximum  row index]
[ec392464]720
[6af2785]721     Example: radius [2 : 5] -3 * scale [2 : 5]
[ec392464]722
[6af2785]723  2) if only some *values* of a given column are needed but the range between
724     the first row and the last row used is not continuous, write this type of
725     expression
[ec392464]726
[6af2785]727     column_name1 [minimum row index1 :  maximum  row index1] , column_name1
728     [minimum row index2 :  maximum  row index2]
[ec392464]729
[6af2785]730     Example: radius [2 : 5] , radius [10 : 25]
[5005ae0]731
[9d93c37]732.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
733
734Combined Batch Fit Mode
735-----------------------
736
[e6c74e8]737The purpose of the Combined Batch Fit is to allow running two or more batch
738fits in sequence without overwriting the output table of results.  This may be
739of interest for example if one is fitting a series of data sets where there is
740a shape change occurring in the series that requires changing the model part
741way through the series; for example a sphere to rod transition.  Indeed the
742regular batch mode does not allow for multiple models and requires all the
743files in the series to be fit with single model and set of parameters.  While
744it is of course possible to just run part of the series as a batch fit using
745model one followed by running another batch fit on the rest of the series with
746model two (and/or model three etc), doing so will overwrite the table of
747outputs from the previous batch fit(s).  This may not be desirable if one is
748interested in comparing the parameters: for example the sphere radius of set
[05b0bf6]749one and the cylinder radius of set two.
[e6c74e8]750
751Method
752^^^^^^
[9d93c37]753
[e6c74e8]754In order to use the *Combined Batch Fit*, first load all the data needed as
755described in :ref:`Loading_data`. Next start up two or more *BatchPage* fits
756following the instructions in :ref:`Batch_Fit_Mode` but **DO NOT PRESS FIT**.
757At this point the *Combine Batch Fit* menu item under the *Fitting menu* should
758be active (if there is one or no *BatchPage* the menu item will be greyed out
759and inactive).  Clicking on *Combine Batch Fit* will bring up a new panel,
760similar to the *Const & Simult Fit* panel. In this case there will be a
761checkbox for each *BatchPage* instead of each *FitPage* that should be included
762in the fit.  Once all are selected, click the Fit button on
763the *BatchPage* to run each batch fit in *sequence*
[9d93c37]764
765.. image:: combine_batch_page.png
766
[e6c74e8]767The batch table will then pop up at the end as for the case of the simple Batch
768Fitting with the following caveats:
[9d93c37]769
770.. note::
771   The order matters.  The parameters in the table will be taken from the model
[e6c74e8]772   used in the first *BatchPage* of the list.  Any parameters from the
773   second and later *BatchPage* s that have the same name as a parameter in the
774   first will show up allowing for plotting of that parameter across the
775   models. The other parameters will not be available in the grid.
[9d93c37]776.. note::
777   a corralary of the above is that currently models created as a sum|multiply
778   model will not work as desired because the generated model parameters have a
779   p#_ appended to the beginning and thus radius and p1_radius will not be
780   recognized as the same parameter.
[5005ae0]781
[9d93c37]782.. image:: combine_batch_grid.png
783
[e6c74e8]784In the example shown above the data is a time series with a shifting peak.
785The first part of the series was fitted using the *broad_peak* model, while
786the rest of the data were fit using the *gaussian_peak* model. Unfortunately the
787time is not listed in the file but the file name contains the information. As
788described in :ref:`Grid_Window`, a column can be added manually, in this case
[5005ae0]789called time, and the peak position plotted against time.
[9d93c37]790
791.. image:: combine_batch_plot.png
792
[05b0bf6]793Note the discontinuity in the peak position.  This reflects the fact that the
794Gaussian fit is a rather poor model for the data and is not actually
[e6c74e8]795finding the peak.
[ec392464]796
797.. ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
[6af2785]798
[84ac3f1]799.*Document History*
800
801| 2017-09-10 Paul Butler
802| 2017-09-15 Steve King
803| 2018-03-05 Paul Butler
Note: See TracBrowser for help on using the repository browser.