<< vare VAR estimations vargranger >>

Grocer >> VAR estimations > varf

varf

VAR forecasting

CALLING SEQUENCE

rvarf=varf(namevar,hprev,arg1,...,argn)

PARAMETERS

Input

* namevar = the tlist results resulting from a VAR, an ECM, a BVAR or a BECM estimation or the same between quotes (to have it's name saved)

* hprev = the prevision period which can be either

  - a [n1 n2] constant vector where n1 and n2 are the lead over the last period of the estimation (n1<=0 means that the forecast begins within the estimation period)

  - a n constant which is equivalent to [1 n] (forecast begins just after the estimation period)

  - a [n1 n2] string vector where n1 and n2 are the time periods for forecasting (a posibility open only if the VAR has been performed with ts)

  - a n string which is equivalent to [1 n] (forecast begins just after the estimation period)

* argi = an argument which can be

  -xp=xx' (necessary if the variables in the VAR has been given as a matrix) where xx is [# of forecasting dates ; # of exogenous variables in the var] for the values of exogenous variables over the forecasting period

  - the string 'noprint' if the user doesn't want to print the results of the forecast

 

Output

* rvarf = a results tlist with:

  . rvarf('meth')  = 'rvarf'

  . rvarf('rvar')  = results tlist of the originating var

  . rvarf('prev')  = matrix of forecasts

  . rvar('prev_namex')  = vector or ts of forecasts called by their names preceded by 'prev_'

DESCRIPTION

Performs forecasting by the mean of a vector autogressive estimation and, by default, displays on screen the estimation results. If the var structure comes from an ecm estimation, then what is stored is the forecasted levels; if it comes from an ordinary var, then what is stored is the forecasted values of the endogenous variables, as they have been given by the user (for instance, if she has given one as del(ts), then what is forecasted is del(ts), not ts).

EXAMPLE

bounds('1982m7','1994m12');
illinos=reshape(illinos,'1982m1')
indiana=reshape(indiana,'1982m1')
kentucky=reshape(kentucky,'1982m1')
michigan=reshape(michigan,'1982m1')
ohio=reshape(ohio,'1982m1')
pennsyvlania=reshape(pennsyvlania,'1982m1')
tennesse=reshape(tennesse,'1982m1')
westvirginia=reshape(westvirginia,'1982m1')
varjls=VAR(2,'endo=illinos,indiana,kentucky,michigan,ohio,pennsyvlania,tennesse,westvirginia','noprint');
rf=varf(varjls,'1995m12')
// Example taken from function varf_d().
// Forecast is done from the first out-of-sample period (1995m1) until 1995m12.
ecmjls=ecm(2,'illinos','indiana','kentucky','michigan','ohio','pennsyvlania','tennesse','westvirginia','noprint');
rf=varf('ecmjls',[-1,12])
// Now, forecast is done from the last but one period (1994m11) until the 12th out-of-sample period (1995m12).

AUTHOR

Eric Dubois 2002

Report an issue
<< vare VAR estimations vargranger >>