GitLab

initval_file and histvfal_file should be more flexible and have functionalities similar to option datafilein estimation

Usage

initval_file

• is used only in perfect foresight
• provides a guess value for computing the solution
• in absence of histval or histval_file provides the initial conditions for the simulation (we don't want to require users to upload two files for initial conditions and guess values

histval_file

• is used for stochastic and perfect foresight models
• provides initial conditions

Current implementation

initval_file

• accept .m, .mat, .xls(x), .csv files
• modifies directly oo_.endo_simul, oo_.exo_simul, but not oo_.exo_det_simul

histval_file

• accept only files prepared by smoother2histval that uses odd format
• modifies directly oo_.endo_simul, oo_.exo_simul, but not oo_.exo_det_simul
• note that histval sets a dseries object

Proposal

1. have the same interface for initval_file and histval_file
2. add an option dseries to be able to pass a dseries object
3. handle references to varexo_det variables
4. use the dseries interface to read files
5. initval_file and histval_file```set dseries`` object
6. modify make_y_ make_ex_ accordingly
7. handle auxiliary variables in a consistent manner (see #1004)

Breaking changes

1. Documented behavior of initval_file: None
2. Documented behavior of histval_file: None
3. Documented behavior of perfect_foresight_setup: the format of files designated in option filename has changed
4. Function initvalf.m
   • M_ and options are added to inputs
   • the old filename input triggers an error
   • function initvalf() sets a result dseries on output
5. Function histvalf.m
   • M_ and options are added to inputs
   • the old filename input triggers an error
   • function histvalf() sets a result dseries on output
   • histvalf.mdoesn't recognize the old smoother2histval file format anymore.
6. Function smoother2histval.m
   • when an output file is requested smoother2histval() creates a .mat file containing a single dseries
   • smoother2histval.m saves now M_.orig_maximum_lagobservations instead ofoo_.maximum_lag`` in order to be able to recompute auxiliary variables
   • It will not be possible to use a file saved with an previous version of smoother2histval

Questions

1. Is it a problem to have an optional dseries input in an instruction called initval_file and no file strictly speaking of?

WORK IN PROGRESS