STAMP 9: enhancements and solved problems

Many new features have been introduced in version 9 of STAMP. The most notable are:

• Various bug fixes on the handling of multivariate structural time series model, including the storage in the OxMetrics database.
• An OxMetrics database with multiple time series can be analyzed sequentially. Each variable is treated by the pre-specified univariate structural time series model. The output after this sequential analysis includes a review of the estimation results and a graphical output summary. The user can access the extracted signals and other graphical output in the usual way using the regular menus.
• Further, the sequential univariate analysis mode produces a summary of results for time series variables that have failed tests or have a goodness-of-fit R$^2$ value lower than 0.05.
• Finally, the sequential univariate analysis mode also produces graphical and written output that indicate how much of the residual covariance matrix can be explained by a principal component analysis (PCA).
• Ox code (Alt-o) generation is updated and extended, for example it includes options for forecasting.

STAMP 8.3: new features

• STAMP 8.3 works under OxMetrics 6.1.
• The Ox code generator is introduced and fully supported by STAMP. This new facility can generate Ox code for the model that is estimated in STAMP. It complements the Batch code generator in STAMP. It is particularly useful for those who use Ox for time series analysis in a production environment.
• The online help facility of STAMP is updated. In particular, the online help for the Batch language and the new Ox code generator are rewritten.

STAMP 8.3: solved problems

• All weights and related computations in the Test/Weights dialog can be carried out, also for time series with missing data.
• The Write forecasts option is combined with a Store forecasts option in the Test/Forecasting dialog. The observations forecasts are stored after confirmation as a new variable with the forecasts attached at the end of the sample. When necessary, the database sample is automatically extended such that the forecast window is included. The in-sample values of the new variable are the same as in the original series.
• The Edit/Save forecasts option in the Test/Forecasting dialog is reactivated for model without explanatory variables.
• The Batch code options for Forecasting is extended; see Batch documentation.
• Variables and components in the Batch code need to be written between accolades. Specifically, in the setcmp batch command we have "level", "slope", "seasonal", "cycle", "ar" and "irregular".
• Inclusion of lagged dependent variables is discouraged. A new facility will be built in for the next version. In this version it is best to treat and to have it as an exogenous variable.

STAMP 8.2: new features

• STAMP 8.2 works under OxMetrics 6.
• The algorithm for the automatic detection of outliers and breaks (option in the "Select components" dialog) is further advanced and works more effectively.

STAMP 8.2: solved problems

• Output of variable names in "Regression effects in final state" is corrected when Interventions are included in the model.
• Regression effects are computed correctly when lagged explanatory variables are included in the model.
• Forecasting graphs are corrected for multivariate models (no multiple lines in graph).
• Convergence criteria now depend on the number of dependent variables in the model.
• Convergence criteria for EM method is adjusted for certain model settings.
• Maximum likelihood estimation of the variance matrices for the components is adjusted for multivariate models when missing values are present.

STAMP 8.1: solved problems

• Solved error in regression output when using time-varying coefficients.
• Solved batch error when replicating current session of STAMP.
• Optimized automatic intervention selection (simultaneous for level and slope).
• Solved error with missing values in graphs for forecasting.
• Fixed forecasting mse with intervention in level.
• Fixed multiple cycle error in estimation.
• Added output for forecasts of components in model.

STAMP 8.0: new features

Many new features have been introduced in version 8 of STAMP. The most notable are:

• Multivariate models

  The multivariate structural time series model where the unobserved components become vectors and the disturbance variances become disturbance variance matrices can be considered for the analysis of a set of multiple time series. The number of multivariate options has increased considerably compared to earlier versions of the program:

  • Select components by equation: Different components can be selected for different equations. This enables the user to analyse time series with different dynamic characteristics jointly. For example, consider two time series where one series may be subject to seasonal dynamics while the other series does not require a seasonal component. The trends of the two time series may move together. STAMP 8 allows the user to select a seasonal component for the first series but not for the second series. This applies to all components in STAMP: trend, seasonal, cycle, autoregressive, irregular, time-varying regressions, etc.
  • Select regressions and interventions by equation: An option for selecting different explanatory variables and interventions for different equations has been available in STAMP versions 5 and 6. However, the current facility of distributing explanatory variables over different equations has improved and is more flexible.
  • Design a dependence structure for each component: Multivariate models in STAMP 5 and 6 were limited in their choice of variance matrices: only full variance matrices of different ranks could be considered. A reduced-rank variance matrix implies common features in multiple time series. This option remains in STAMP but the specification has changed slightly. The disturbance variance matrix imposes a dependence structure within the component vector (between the different equations). This dependence can be designed by the user in a simple way and for each component separately. For example, the cycle component in equation 1 can be forced to depend on the cycles in equations 2 and 3 only.
  • In STAMP 8 different variance matrices for different disturbances can be chosen: The range of variance matrices includes scalar and diagonal matrices, scaled matrices of ones (when applied to the slope component, this implies balanced growth) and one rank plus diagonal matrices. The latter case implies that a vector component can be decomposed into common and idiosyncratic effects. In many applications, these different specifications can be interpreted easily and can be highly interesting.
  • The multivariate options extend to all models introduced in STAMP 7: This includes the higher-order smooth trend models, the higher-order (bandpass) cycle components and the (vector) autoregressive components of orders 1 and 2.
  • Missing observations: They can also be handled within multivariate time series models. This allows the interpolation of missing observations through time but also through different time series.
  • Forecasting of multivariate time series made simple: In particular, STAMP 8 allows the incorporation of available future observations for the explanatory variables in the database. Furthermore, future observations of dependent variables are considered in graphical presentations of forecasts and for the measurement of forecast accuracy (using standard measures such as the root mean squared forecast error (RMSE) and the mean absolute percentage error (MAPE).
  • Estimation of parameters in multivariate time series models is based on exact procedures: The diffuse initialisation of the Kalman filter is implemented, the exact likelihood function is computed and the score function with respect to variance parameters is computed analytically and fast. This leads to a robust estimation procedure in STAMP 8 and a relatively fast estimation process.
  • The number of graphical output for multivariate models is increased: STAMP 8 offers an easy handling of the graphical output. An option for graphics output selection for each equation is provided. The powerful tools in OxMetrics 5 to edit graphical output are fully available to STAMP 8 users.

• Automatic outlier and break detection

  Another major development in STAMP 8 is the implementation of a new automatic detection procedure for outliers and breaks in univariate and multivariate time series models. The following features are available:

  • STAMP 8 is able to propose a set of potential outliers and trend breaks for univariate and multivariate time series. It is a basic but effective two-step procedure based on the auxiliary residuals. First the selected model is estimated and the diagnostics are investigated. Then a first (larger) set of potential outliers and trend breaks are selected from the auxiliary residuals. After re-estimation of the model, only those interventions survive that are sufficiently significant. In the multivariate case, this selection procedure is carried out jointly for each equation in the model.
  • After the automatic selection, the results are reported. All considered outliers and breaks are kept in the intervention dialog and they can be deleted from the model or added to the model in the usual way and implemented as in STAMP 7. For future use, the interventions can be saved. It prevents the manual input of outliers and breaks altogether.
  • The automatic selection procedure can be repeated with the inclusion of a fixed set of explanatory and intervention variables.

• Other new features
  • Each parameter in the models of STAMP 8 can be edited directly. Parameters can be kept fixed at a particular value. Variances can be kept fixed at values relative to a particular variance of another component (q-ratio). This facility also applies to multivariate models.
  • General forecasting options have been extended and made more flexible. The number of output options for prediction and forecasting have been increased. Future values of explanatory variables available in the database can be used for the forecasting of dependent variables.
  • More output diagnostics are presented for predictions (one-step and multi-step), auxiliary residuals and weight and gain functions.