Back to TotalFit Methods

Choosing ranges for Monte-Carlo runs

Setting ranges for starting parameters (Monte-Carlo ranges)

One may choose any ranges to start fitting from to see if there are local minima elsewhere in a parameter space. However, this feature only works well with relatively simple functions. Fitting with complex functions may not be able to converge with too extreme combination of parameters. When trying different ranges it is useful to turn monte_carlo_monitoring='SEE-ALL-FIGURES' so you see how fits look and how simulated data look.  See docs/concepts/Concepts of data fitting/Diagnostics of convergence in fitting runs for more information on how to control and troubleshoot Monte-Carlo runs and use of vary_starting_parameters property.

 

Setting absolute limits on fitting parameters

Absolute limits are very advisable to have: they restrict most fitting algorithms (except ones using fmeansearch as a basic algorithm; 'simplex'---in TotalFit convention) to meaningful ranges and reduce fitting time because it may take most time for trial to converge if the optimization parameters venture too far from the optimum.

 

For a specific usage see TotalFit.m