model_selection

Synopsis

Select the best model from a set of optimised models.

Defaults

model_selection(method=`AIC', modsel_pipe=None, bundle=None, pipes=None)

Keyword arguments

method: The model selection technique (see below).

modsel_pipe: The name of the new data pipe which will be created by this user function by the copying of the selected data pipe.

bundle: The optional pipe bundle is a special grouping or clustering of data pipes. If this is specified, the newly created data pipe will be added to this bundle.

pipes: An array containing the names of all data pipes to include in model selection.

Description

The following model selection methods are supported:

AIC -

Akaike's Information Criteria.

AICc -

Small sample size corrected AIC.

BIC -

Bayesian or Schwarz Information Criteria.

Bootstrap -

Bootstrap model selection.

CV -

Single-item-out cross-validation.

Expect -

The expected overall discrepancy (the true values of the parameters are required).

Farrow -

Old model-free method by Farrow et al., 1994.

Palmer -

Old model-free method by Mandel et al., 1995.

Overall -

The realised overall discrepancy (the true values of the parameters are required).

For the methods `Bootstrap', `Expect', and `Overall', the Monte Carlo simulations should have previously been executed with the monte_carlo.create_data method set to Bootstrapping to modify its behaviour.

If the data pipes have not been specified, then all data pipes will be used for model selection.

Prompt examples

For model-free analysis, if the preset models 1 to 5 are minimised and loaded into the program, the following commands will carry out AIC model selection and to place the selected results into the `mixed' data pipe, type one of:

[numbers=none]
relax> model_selection('AIC', 'mixed')

[numbers=none]
relax> model_selection(method='AIC', modsel_pipe='mixed')

[numbers=none]
relax> model_selection('AIC', 'mixed', ['m1', 'm2', 'm3', 'm4', 'm5'])

[numbers=none]
relax> model_selection(method='AIC', modsel_pipe='mixed', pipes=['m1', 'm2', 'm3', 'm4', 'm5'])