What is G7?¶
G7 is the regression analysis member of a closely integrated package of programs for building economic models. It can be used to look at data, to estimate a single equation or a system of equations, to build a simultaneous equation model, or to create a multisectoral model involving hundreds of equations. It has extensive capabilities for:
Making, using, and exploring data banks. Over twenty different banks can be assigned at once. Data can be shown in graphs or in grids resembling spreadsheets.
Rapidly drawing and annotating graphs and saving them so that they can be imported into documents.
Producing tables of data and modeling results.
Editing and displaying text files.
Combining and transforming variables. Transformations available include algebraic, exponential, logarithmic, sine, cumulation with decay, interpolation, change of frequency, linking and benchmarking of series, cutting off negatives, filtering, and converting all positive elements to 1.
Looping, such as over sector numbers and titles.
Performing regressions by ordinary least squares, least squares with soft constraints and distributed lags, stacked regression with constraints across equations, pooled regression, SUR, 2SLS, 3SLS, recursive regression, ARIMA, and non-linear least squares. Regression results include – besides the usual regression coefficients and t-values, \(R^{2}\), \(\rho\), and DW statistics – for each variable its marginal explanatory value, elasticity at the mean, beta coefficient, and the normalized sum of squared residuals after its introduction. F statistics for groups of variables are available, as is the leverage variable for finding outlying observations. If requested, the matrix of simple correlations among all the variables, the derivatives of regression coefficients with respect to one another, and the variance-covariance matrix of the regression coefficients can be shown. Statistics for testing whether the residuals are normal are available. Information necessary for stochastic simulation can be saved. Missing values in any variable are noted.
Building simultaneous equation macro-economic models. An enormously useful feature absent in other programs is the ability to create exogenous variables, such as tax rates, conveniently at the point at which they are used. The building process checks the model for inconsistencies in the definition of variables. A simple check on the correctness of the model’s identities is provided. Trend projections of all exogenous variables can be generated automatically. The building process writes a C program for solving the model. This program then is automatically compiled and linked, usually in a few seconds. It then operates extremely rapidly. Also, any calculations not easily expressed in G7 commands can be accomplished by C code that is passed through to the program for solving the model. Models easily can be run with various sorts of modifications to the results of the regression equations. Stochastic simulations can be run.
An objective function, specified by the model builder, can be optimized relative to selected regression coefficients. If the objective function relates to how well the model reproduces history, this ability can be viewed as an advanced econometric method for fitting equations. If it is a measure of welfare, it can be viewed as a method of design of optimal policy.
Building interindustry models or other models that require extensive matrix and vector operations. Such models can include a linked system of national or regional models or a model of bilateral world trade at a detailed commodity level. Results of these models can be shown in grids resembling spreadsheets. Optimization is also available for these models.
G7 Versions 7.0 and higher require the Windows 2000 or newer operating system.
The first portion of this Help system is a User Guide intended to be read in order by clicking on the Next and Previous links on the Help reader form. The second portion is a Reference Manual.
