Differences between revisions 13 and 14
Revision 13 as of 2011-03-14 14:34:30
Size: 9275
Editor: JakeSmith
Comment:
Revision 14 as of 2011-03-14 20:36:32
Size: 12434
Editor: JakeSmith
Comment:
Deletions are marked like this. Additions are marked like this.
Line 7: Line 7:
|| '''Flag''' |||| '''Argument''' |||| '''Comments''' ||
|| {{{irf}}} |||| <<integer>> |||| Produce Impulse Response Graphs with horizon=integer ||
|| {{{bayesian_irf}}} |||| -- |||| Produce Bayesian Impulse Response Graphs with horizon=irf ||
|| {{{forecast}}} |||| <<integer>> |||| Produce Forecast (banded?) with horizon=integer ||
|| {{{coefficients_prior_hyperparameters}}} |||| <<vec_double>> |||| Sets the hyper parameters for the sbvar (mu in sz_prd.m), this should be a vector of doubles of length 6 ||
|| {{{mode_compute}}} |||| <<integer>> |||| Default 1 Uses csminwel to estimate the model, mode_compute=0 will not try to estimate the model ||
|| {{{mode_file}}} |||| <<string>> |||| Mode file to load instead of re-estimating the model ||
|| {{{mh_replic}}} |||| <<integer>> |||| Renaming n_draws* to match dynare convention ||
|| {{{drop}}} |||| <<integer>> |||| Renaming burning* to match dynare convention ||
|| {{{thinning_factor}}} |||| <<integer>> |||| total draws made = thinning_factor*mh_replic + drop, this should default to 1||
|| {{{adaptive_mh_draws}}} |||| <integer> |||| Renaming mh* to be more descriptive ||
|| '''Flag''' |||| '''Argument''' |||| '''Comments''' |||| '''Default''' |||| '''Variable Name''' |||| '''Previous Name''' ||
|| {{{mode_compute}}}* |||| <<integer>> ||||~- '''1.''' Use csminwel to estimate the mode '''0.''' Will not try to estimate the model -~|||| 1 |||| |||| estimate_msmodel ||
|| {{{mode_file}}} |||| <<string>> ||||~- Mode file to load instead of re-estimating the model -~|||| |||| |||| ||
|| {{{compute_mdd}}}* |||| <<integer>> |||| ~- '''1.''' Compute marginal data densities '''0.''' Otherwise -~|||| 1 |||| |||| compute_mdd ||
|| {{{compute_probabilities}}}* |||| <<integer>> |||| ~- '''1.''' Compute probabilites '''0.''' Otherwise -~|||| 1 |||| |||| compute_probabilities ||
|| {{{irf}}} |||| <<integer>> ||||~- Produce Impulse Response Graphs with horizon=integer -~|||| |||| |||| ||
|| {{{bayesian_irf}}} |||| -- ||||~- Produce Bayesian Impulse Response Graphs with horizon=irf -~|||| |||| |||| ||
|| {{{forecast}}} |||| <<integer>> ||||~- Produce Forecast with horizon=integer -~|||| |||| |||| ||
|| {{{coefficients_prior_hyperparameters}}} |||| <<vec_double>> ||||~- Sets the hyper parameters for the sbvar (mu in sz_prd.m), this should be a vector of doubles of length 6 -~|||| [1.0 1.0 0.1 1.2 1.0 1.0] |||| |||| ||
|| {{{mh_replic}}}* |||| <<integer>> ||||~- The number of draws to save from the simulation -~|||| 1000|||| n_draws |||| n_draws ||
|| {{{drop}}} |||| <<integer>> ||||~- The number of samples to drop at the beginning of the sampling period -~|||| |||| |||| ||
|| {{{thinning_factor}}}* |||| <<integer>> ||||~- total draws = ( thinning_factor x mh_replic ) + drop -~|||| 1 |||| |||| thinning_factor ||
|| {{{adaptive_mh_draws}}} |||| <<integer>> ||||~- Number of draws for the Adaptive MH step -~|||| |||| |||| ||
|| {{{load_mh_file}}} |||| <<string>> ||||~- Filename of previously computed simulation run, load this instead of simulating the model again. -~|||| |||| |||| ||
|| {{{freq}}}* |||| <<integer>> ||||~- quarters or months -~|||| 4 |||| q_m |||| freq ||
|| {{{initial_year}}}* |||| <<integer>> ||||~- beginning of the year -~|||| |||| yrBin |||| initial_year ||
|| {{{initial_subperiod}}}* |||| <<integer>> |||| ~-beginning of the quarter or month-~ |||| 1 |||| qmBin |||| initial_subperiod ||
|| {{{final_year}}}* |||| <<integer>> |||| ~-final year-~ |||| |||| yrFin |||| final_year ||
|| {{{final_subperiod}}}* |||| <<integer>> ||||~- final month or quarter-~ |||| |||| qmFin |||| final_subperiod ||
|| {{{data_file}}}* |||| <<string>> ||||~- Load data-~ |||| |||| |||| data_file ||
|| {{{varlist}}}* |||| <<vec_value>> ||||~- name of variables -~|||| |||| varlist |||| varlist ||
|| {{{restriction_fname}}}* |||| <<string>> |||| ~-file for restriction on time variations-~ |||| |||| idfile_const |||| restriction_fname ||
|| {{{nlags}}}* |||| <<integer>> ||||~- Number of lags -~|||| 4 |||| lags |||| nlags ||
|| {{{cross_restrictions}}}* |||| <<integer>> ||||~- '''1''': cross-A0-and-A+ restrictions <<BR>> '''0''': idfile_const is all we have-~ |||| 0 |||| indxc0Pres |||| cross_restrictions ||
|| {{{contemp_reduced_form}}}* |||| <<integer>> |||| ~-'''1''': contemporaneous recursive reduced form <<BR>> '''0''': restricted (non-recursive) form -~|||| 0 |||| Rform |||| comtemp_reduced_form ||
|| {{{bayesian_prior}}}* |||| <<integer>> |||| ~-'''1''': Bayesian prior <<BR>> '''0''': no prior-~ |||| 1 |||| indxPrior |||| bayesian_prior ||
|| {{{alpha}}}* |||| <<double>> ||||~-Alpha value for squared time-varying structural shock lambda.-~ |||| 1.0 |||| galp |||| alpha ||
|| {{{beta}}}* |||| <<double>> ||||~-Beta value for squared time-varying structural shock lambda.-~ |||| 1.0 |||| gbeta |||| beta ||
|| {{{gsig2_lmd}}}* |||| <<integer>> ||||~-Case 4 (no state change across variables (i) and across lags (l); only one scale factor for all lagged variables change from state to state). Normal prior variance for structural shock lambda, one scale factor for all lagged D+ in a given equation.-~ ||||50^2^ |||| gsig2_lmd |||| gsig2_lmd ||
|| {{{gsig2_lmdm}}}* |||| <<integer>> ||||~-Case 3 (no state change across lags (l) but allows all variables for a given lag to switch states). Normal prior variance for different variables in lagged D+.-~ ||||50^2^ |||| gsig2_lmdm |||| gsig2_lmdm ||

NB: Below, an * indicates that the option already exists and is referenced in MarkovSwitchingOptions.

ms_sbvar new/renamed options

Flag

Argument

Comments

Default

Variable Name

Previous Name

mode_compute*

<<integer>>

1. Use csminwel to estimate the mode 0. Will not try to estimate the model

1

estimate_msmodel

mode_file

<<string>>

Mode file to load instead of re-estimating the model

compute_mdd*

<<integer>>

1. Compute marginal data densities 0. Otherwise

1

compute_mdd

compute_probabilities*

<<integer>>

1. Compute probabilites 0. Otherwise

1

compute_probabilities

irf

<<integer>>

Produce Impulse Response Graphs with horizon=integer

bayesian_irf

--

Produce Bayesian Impulse Response Graphs with horizon=irf

forecast

<<integer>>

Produce Forecast with horizon=integer

coefficients_prior_hyperparameters

<<vec_double>>

Sets the hyper parameters for the sbvar (mu in sz_prd.m), this should be a vector of doubles of length 6

[1.0 1.0 0.1 1.2 1.0 1.0]

mh_replic*

<<integer>>

The number of draws to save from the simulation

1000

n_draws

n_draws

drop

<<integer>>

The number of samples to drop at the beginning of the sampling period

thinning_factor*

<<integer>>

total draws = ( thinning_factor x mh_replic ) + drop

1

thinning_factor

adaptive_mh_draws

<<integer>>

Number of draws for the Adaptive MH step

load_mh_file

<<string>>

Filename of previously computed simulation run, load this instead of simulating the model again.

freq*

<<integer>>

quarters or months

4

q_m

freq

initial_year*

<<integer>>

beginning of the year

yrBin

initial_year

initial_subperiod*

<<integer>>

beginning of the quarter or month

1

qmBin

initial_subperiod

final_year*

<<integer>>

final year

yrFin

final_year

final_subperiod*

<<integer>>

final month or quarter

qmFin

final_subperiod

data_file*

<<string>>

Load data

data_file

varlist*

<<vec_value>>

name of variables

varlist

varlist

restriction_fname*

<<string>>

file for restriction on time variations

idfile_const

restriction_fname

nlags*

<<integer>>

Number of lags

4

lags

nlags

cross_restrictions*

<<integer>>

1: cross-A0-and-A+ restrictions
0: idfile_const is all we have

0

indxc0Pres

cross_restrictions

contemp_reduced_form*

<<integer>>

1: contemporaneous recursive reduced form
0: restricted (non-recursive) form

0

Rform

comtemp_reduced_form

bayesian_prior*

<<integer>>

1: Bayesian prior
0: no prior

1

indxPrior

bayesian_prior

alpha*

<<double>>

Alpha value for squared time-varying structural shock lambda.

1.0

galp

alpha

beta*

<<double>>

Beta value for squared time-varying structural shock lambda.

1.0

gbeta

beta

gsig2_lmd*

<<integer>>

Case 4 (no state change across variables (i) and across lags (l); only one scale factor for all lagged variables change from state to state). Normal prior variance for structural shock lambda, one scale factor for all lagged D+ in a given equation.

502

gsig2_lmd

gsig2_lmd

gsig2_lmdm*

<<integer>>

Case 3 (no state change across lags (l) but allows all variables for a given lag to switch states). Normal prior variance for different variables in lagged D+.

502

gsig2_lmdm

gsig2_lmdm

Create Init File Mex Function

Create Init File Syntax (from switch_dw/state_space/sbvar/create_init_file.c):

create_init_file <matlab filename> <markov filename> <file tag>

Standalone Argument

Dynare Option Name

Comments

<matlab filename>

None. Set automatically in sz_prd.m

matlab file name

<markov filename>

markov_file*

Markov file name

<file tag>

output_file_tag*

The output file tag

Estimation, Simulation, Probabilities, Marginal Data Densitiy Mex Function

Top Level Options (from switch_dw/state_space/sbvar/dw_sbvar_command_line.c):

Standalone Flag

Argument

Standalone Default

Dynare Option Name

Comments

estimate

--

estimate_msmodel*

produces estimation - posterior mode or maximum likelihood

simulate

--

print_draws*

produces simulation run

probabilities

--

compute_probabilities*

produces smoothed or filtered probabilities

mdd

--

compute_mdd*

produces marginal data density from posterior draws

forecast

--

produces forecast

ir

--

produces impulse responses

historical

--

produces historical decomposition

smoothedshocks

--

produces smoothed shocks

smoothedstates

--

produces smoothed states

seed

<<integer>>

0 (i.e. random)

set by set_dynare_seed() in .mod file

seed value for random number generation


General Options (from switch_dw/state_space/sbvar/dw_sbvar_command_line.c):

Standalone Flag

Argument

Standalone Default

Dynare Option Name

Comments

di

<<string>>

./

not supported

input files are in specified directory

do

<<string>>

./

not supported

output files are in specified directory

fs

<<string>>

specification file name

fp

<<string>>

parameter file name

ph

<<string>>

header for the parameters

pho

<<string>>

parameter header used for output

MLE

<<string>>

Determines if estimation finds the posterior mode or maximum likelihood. If ph or is not present, determines whether the input parameter header is "MLE: " or "Posterior mode: ".

fo

<<string>>

output file name

ft

<<string>>

The input file tag. Is used to create input filenames if the fs or fp options are not present.

fto

<<string>>

associated with ft

output_file_tag*

The output file tag. Used to create output filenames.


Estimation Options (from switch_dw/state_space/sbvar/sbvar_estimate.c):

Standalone Flag

Argument

Standalone Default

Dynare Option Name

Comments

cb

<<double>>

1.0e-03

convergence criterion starting value

ce

<<double>>

1.0e-06

convergence criterion ending value

ci

<<double>>

0.1

convergence criterion increment value

ib

<<integer>>

50

maximum interations starting value

ii

<<double>>

2.0

maximum interations increment value

mb

<<integer>>

100

maximum block iterations


Simulation Options (from switch_dw/state_space/sbvar/sbvar_simulate.c):

Standalone Flag

Argument

Standalone Default

Dynare Option Name

Comments

ndraws

<<integer>>

1000

n_draws*

Number of draws to save

burnin

<<integer>>

0.1*ndraws

draws_nbr_burn_in
two similar options existed, but were appended with _1 and _2

Number of burn-in draws

thin

<<integer>>

1

thinning_factor*

Thinning factor. Total number of draws made is thin*ndraws+burnin

mh

<<integer>>

30000

Tuning period for Metropolis-Hasting draws

flat

--

off

Produce flat output file and header

nofree

--

off

Do not produce free parameters file

nd1

--

off

Normalize diagonal of A0 to one (flat output only)


Probabilities Options (from switch_dw/state_space/sbvar/sbvar_probabilities.c):

Standalone Flag

Argument

Standalone Default

Dynare Option Name

Comments

filtered

--

off

Compute filtered probabilities

smoothed

--

on if neither filtered nor real_time_smoothed is passed

Compute smoothed probabilities

real_time_smoothed

--

off

Compute smoothed probabilities based on time t information for 0 <= t <= nobs


Marginal Data Density Options (from switch_dw/switching/dw_mdd_switch.c):

Standalone Flag

Argument

Standalone Default

Dynare Option Name

Comments

t

<<string>>

filemname tag

ft

<<string>>

filemname tag -- supersedes t

outtag

<<string>>

filemname tag -- supersedes t and ft

pf

<<string>>

simulation_<tag>.out

posterior draws fileanme

pt

<<integer>>

2

proposal type (1=gaussian, 2=power, 3=truncated power, 4=step, 5=truncated gaussian)

d

<<integer>>

100,000

number proposal draws

use_mode

--

on

use posterior mode as center

use_mean

--

off

use posterior mean as center

of

<<string>>

mdd_<tag>.out

output filename

l

<<double>>

0.1

Lower cutoff for in terms of probability. Used by truncated power, truncated gaussian, step, and uniform.

h

<<double>>

0.9

Upper cutoff for in terms of probability. Used by power, truncated power, truncated gaussian, step, and uniform.

graph

--

off

Produce output suitable for graphing the marginal cummulative densities of the radius of the posterior and elliptical distributions. The marginal data density is not computed.

print

--

off

Prints the matrices posterior and proposal

DynareWiki: MSSbvarOptions (last edited 2012-03-08 14:53:45 by HoutanBastani)