Skip to content

ModelSelection.md

Latest commit

 

History

History
77 lines (50 loc) · 5.38 KB

ModelSelection.md

File metadata and controls

77 lines (50 loc) · 5.38 KB

Back to Fitting

Model selection

The ModelSelector module

The fourth step in the modeling run is the selection of best models for a given model category. This task is performed by a specific external module called ModelSelector. This module can be launched by the corresponding function called ModelSelector():

import RTModel
rtm = RTModel.RTModel('/event001')
rtm.Reader()
rtm.InitCond()
rtm.launch_fits('PS')
rtm.ModelSelector('PS')

With this code, we first perform the data pre-processing by Reader, we set the initial conditions by InitCond, we launch all fits of the single-lens-single-source model with launch_fits and then we select the best models within this category with ModelSelector.

In the /event001 directory you will see that the following products appear:

  • A new subdirectory called Models/ is created. This will contain the best models for each category.
  • One or more files named PSXXXX-X.txt containing the details of the selected models. Each model is identified by the label for the model category followed by the number of initial condition and then by the fit number. Any model obtained by user-defined initial conditions will carry the same label specified by the user.
  • In addition, in the InitCond/ subdirectory, some initial conditions files are updated to include more initial conditions obtained by perturbing the best models found in this category. For example, after the single-lens-single-source fits, initial conditions for binary lenses starting from best models found with single lens are added. These are particularly useful to model small anomalies due to planets.

After the execution of ModelSelector, you may call the run() function to complete the modeling run or continue with other calls to launch_fits() and ModelSelector(), or going to final assessment with Finalizer(), depending on your intentions.

After the model selection, the preliminary models calculated by launch_fits will no longer be used. If you do not need to perform any additional fits for the same model category, it is wise to clean up the preliminary models by rtm.cleanup_preliminary_models().

Model files

Each model file contains:

  • The parameters of the model, starting from the non-linear parameters as described in Model categories, followed by the blend and source fluxes for each dataset, in the order shown in FilterToData.txt, four anomaly parameters and finally the chi square.
  • The 1-sigma error for each parameter as listed in the first line, except for the chi square and the anomaly parameters.
  • The covariance matrix for the parameters as used in the fit. Some of them are fit in log scale (see Model categories).

The model selection

The ModelSelector module sorts all preliminary model of the chosen category by their chi square. Models with overlapping covariance ellipsoid are discarded as duplicates. Reflections are considered according to the symmetries of the model category.

Options for model selection

The config_ModelSelector() function

The user may specify his/her own options in model selection by calling the config_ModelSelector() function with the proper options:

import RTModel
rtm = RTModel.RTModel('/event001')
rtm.config_ModelSelector(sigmasoverlap = 3.0, sigmachisquare = 1.0, maxmodels = 10)
rtm.run()

The call to config_ModelSelector() will affect all following executions of the ModelSelector module, whether called through run() or ModelSelector(). If you want to change your options, you may call config_ModelSelector() again.

Description of the options

Here we describe the options for LevMar in detail indicating their default values.

  • sigmasoverlap = 3.0: Maximum number of sigmas to declare overlap.
  • sigmachisquare = 1.0: Besides the best model, ModelSelector retains competing models up to a threshold given by sqrt(2*chisqure), which represents one sigma in the chi square distribution. This threshold can be changed by this option to include less or more competing models in the final selection.
  • maxmodels = 10: Maximum number of competing models to report

All options are also accessible separately as properties of the RTModel class. The user may thus modify the value of each option one by one. The names of the properties are the same as the options in config_ModelSelector() with the prefix ModelSelector_, as shown in the example below:

rtm.ModelSelector_sigmasoverlap = 5.0
rtm.ModelSelector_maxmodels = 2

An empty call to config_ModelSelector() with no parameters will reset all variables to the default values. Notice that the options that are not explicitly specified in the call to config_ModelSelector() are always reset to their default values. This is also true if you previously used the recover_options() function to inherit the options from a previous run (see Archiving and updating).

Recording the options

In each modeling run, the options for ModelSelector are stored in the file ModelSelector.ini in the /ini subdirectory within the event directory for later reference. If the modeling run is archived, also the whole /ini subdirectory is saved so that the user may check the options used in each modeling run. The function recover_options() can be used to load the options from a previous run.

Go to Final assessment