___ ____ ____ ____ ____(R) /__ / ____/ / ____/ ___/ / /___/ / /___/ Statistics/Data Analysis Title svarih postestimation -- Postestimation tools for svarih Description The following postestimation commands are of special interest after svarih: Command Description ------------------------------------------------------------------------- dsimih create and analyze SIRFs and SFEVDs svarih, cmat display estimated coefficients in matrix form ------------------------------------------------------------------------- There is no direct interface to apply VAR postestimation tools to the underlying VAR of svarih. If you want to do this run the VAR separately using var, then apply the VAR postestimation tools. See vargranger, valmar, varnorm, varsoc, varstable, and varwle. Remember that these tools have to be applied in concordance with the assumption of distinct volatility regimes. Forecasts for svarih are not implemented in fcast-type commands. Instead, use predict (see below). The following standard postestimation commands are also available: Command Description ------------------------------------------------------------------------- estat AIC, BIC, VCE, and estimation sample summary estimates cataloging estimation results lincom point estimates, standard errors, testing, and inference for linear combinations of coefficients lrtest likelihood-ratio test nlcom point estimates, standard errors, testing, and inference for nonlinear combinations of coefficients predict fitted values, residuals, shocks, historical decompositions test Wald tests of simple and composite linear hypotheses testnl Wald tests of nonlinear hypotheses ------------------------------------------------------------------------- Syntax for predict predict [type] {newvar|stub} [if] [in] [, statistic {equation(eqno|eqname) llu_suboptions | hd_suboptions} ] statistic description ------------------------------------------------------------------------- xb fitted values; the default residuals residuals shocks shock series hdecomp historical decompositions ------------------------------------------------------------------------- llu_suboptions description ------------------------------------------------------------------------- notnormalized generate shocks whose variances in regime 2 correspond to the elements of the L-matrix rather than being 1 ------------------------------------------------------------------------- hd_suboptions description ------------------------------------------------------------------------- hdshock(eqno|eqname) equation whose shock is examined hdbeg(t1) beginning date of historical decomposition hdend(tN) ending date of historical decomposition ------------------------------------------------------------------------- fitted values, residuals and shocks are available both in and out of sample; type predict ... if e(sample) if wanted only for the estimation sample. Predictions for svarih bac and svarih llu are based on GLS-VAR coefficients if option glsiter(#), #>0, was used, and on VAR coefficients otherwise. Options for predict +------+ ----+ Main +------------------------------------------------------------- xb calculates fitted values. The default. Each call to predict generates one fitted values series. residuals calculates residuals. Each call to predict generates one residual series. Note that the fitted values or residuals generated are the ones from the underlying GLS-VAR or VAR regression(s). Official Stata's svar will not let you create VAR residuals if your svar specification produces an overidentified model. No such check is performed in svarih. In the svar case, you can easily generate VAR residuals by running an equivalent var command and then using predict afterwards. In the svarih case, the generation of correct VAR residuals underlying svarih estimation would be too cumbersome for the user, so predict after svarih will always let you generate residuals. shocks calculates the implied shock series. Each call to predict generates one shock series. hdecomp calculates historical decompositions. Required complementary options are hdbeg and hdend. Option hdshock may optionally by supplied. In contrast to options xb, residuals and shocks, each call to predict generates neqs variables, where neqs is the number of equations in the model. The naming convention for the neqs new variables is stub+eqname, where eqname is the name of a model equation. The data set may not contain a variable named stub. equation(eqno|eqname) is allowed with options residuals and shocks only and specifies the equation to which you are referring. equation() is filled in with one eqno or eqname for options residuals and shocks. equation(#1) would mean that the calculation is to be made for the first equation, equation(#2) would mean the second, and so on. You could also refer to the equation by its name; thus, equation(income) would refer to the equation named income and equation(hours), to the equation named hours. If you do not specify equation(), the results are the same as if you specified equation(#1). +----------------------------------+ ----+ Suboptions for svarih llutkepohl +--------------------------------- notnormalized will calculate series L^(0.5)*e_t, not e_t. For the basic equations of the model, see section Model equations of svarih llutkepohl. +------------------------------------------+ ----+ Suboptions for historical decompositions +------------------------- hdshock(eqno|eqname) uses the calculated shock for the equation corresponding to eqno or eqname in order to calculate the historical decomposition. eqno or eqname can be specified as in option equation. The i-th variable created contains a counterfactual series for the i-th endogenous variable between periods hdbeg and hdend, assuming that only shocks for the equation corresponding to eqno/eqname are present. If you do not specify option hdshock, the 'baseline' forecast, i.e. the regular forecast is calculated. This is identical to forecasts produced by fcast compute. Usually, what is of interest is the difference between the counterfactual of only one shock and the baseline where no shock occurs. In order to create such series, two calls to predict and additional variable subtraction statements are necessary. hdbeg(t1) and hdend(tN) specify the beginning and ending periods of the historical decomposition. t1 and tN can be numeric values that refer to the numeric date encoding of the frequency of the data set. In addition, they can be date strings conformable with the frequency of the data set. Like the option dynamic of fcast compute, option hdbeg of predict creates series starting in period hdbeg-1. In contrast to fcast compute, option hdecomp does not automatically extend the data set if necessitated by option hdend. This can be accommodated easily through a previous tsappend statement. If exogenous variables are in the model, they should have nonmissing values in the date range t1-tN. If a missing value occurs in a certain period, all calculated values subsequent to that period would be missing too which is why predict generates an error in these cases. Historical decompositions currently can only be calculated for daily and lower frequencies. For more information on using predict after multiple-equation commands,