• TODO
  1. Use new GSL 2.X nonlinear fitting methods (requires GSL 2.X)
  2. Add button to create a label with fit results.
  3. support asymmetric errors when fitting. - DIFFICULT errors are currently taken into account by symmetric weights. How can we support asymmetric errors?
  4. Scale the Jacobian when parameters are bounded to get a realistic covariance matrix (i.e. errors). See links of A.
  5. Guess initial parameters (see Origin parameter initializing) - Can only be done for simple fit models. To get good initial values a fit should be done :-) - EVALUATE
  6. Shared parameter of multiple data sets (ensemble averaging (?), see Origin) - Fitting multiple data sets is not supported yet - EVALUATE

• LINKS:
  • http://magicplot.com/wiki/fitting

• DONE:
  • Improved parameter view: https://bugs.kde.org/show_bug.cgi?id=386799
  • Preview with start parameter values (https://bugs.kde.org/show_bug.cgi?id=386800, https://bugs.kde.org/show_bug.cgi?id=394223)

A. contraints on parameter - DONE

• https://lmfit.github.io/lmfit-py/bounds.html
• https://root.cern.ch/doc/master/classTMinuit.html
• http://stackoverflow.com/questions/11312004/python-parameter-transformation-according-to-minuit

B. support fix parameter - DONE

• https://sourceware.org/ml/gsl-discuss/2002-q2/msg00208.html
• https://lists.gnu.org/archive/html/help-gsl/2005-04/msg00013.html
• http://www.physics.brocku.ca/~tharroun/gsl_fit/group__lstsq.html

C. use resulting parameter as new start values (optional) - DONE
D. evaluate full range/selected range of fitted function - DONE
E. Option to ignore errors when fitting. - DONE

F. Find fit parameter in user supplied fit function automatically - DONE

• Currently the fit parameter has to be given by the user which is more work than necessary. We should check if we can get this information from the parser.

G. Improve output of results. - DONE

• use tables to improve visibility - DONE
• calculate t, p- value and conf. interval of parameter - DONE
• calculate p value for chi-square test - DONE
• calculate p value for F test - DONE

• https://reference.wolfram.com/language/howto/FitModelsWithMeasurementErrors.html
• https://de.wikipedia.org/wiki/Einstichproben-t-Test
• http://reliawiki.org/index.php/Multiple_Linear_Regression_Analysis#Estimating_Regression_Models_Using_Least_Squares
• https://onlinecourses.science.psu.edu/stat501/node/297
• http://reliawiki.org/index.php/Simple_Linear_Regression_Analysis
• https://onlinecourses.science.psu.edu/stat501/node/295

H. support different weighting types - DONE

• http://www.originlab.de/doc/Origin-Help/FIt-with-Err-Weight
• http://reference.wolfram.com/language/ref/Weights.html
• http://www.graphpad.com/guides/prism/6/curve-fitting/index.htm?reg_weighted_nonlinear_regression.htm

I. support x-errors for fitting - DONE

• http://wwwhep.physik.uni-freiburg.de/fp/origin/Kap4.pdf
• http://stackoverflow.com/questions/22670057/linear-fitting-in-python-with-uncertainty-in-both-x-and-y-coordinates
• http://stackoverflow.com/questions/12918968/gnuplot-fitting-with-error-variable-in-x-and-y
• https://www.mpp.mpg.de/~caldwell/ss05/Lecture11.pdf
• http://originlab.com/doc/Origin-Help/LinearFit-XErr-Dialog
• http://www.originlab.com/doc/Origin-Help/Ref-Linear-XErr
• https://stats.stackexchange.com/questions/222335/line-of-best-fit-for-data-with-error-in-x-and-y
• http://www.scielo.br/scielo.php?script=sci_arttext&pid=S0100-40422013000600025

J. Implement missing fit models for statistical distributions - DONE