GAUSSX

Desktop Econometric Analysis for GAUSS

[contact addresses][Tips & Tricks][events][homepage]

GAUSSX for Windows is now available. It requires the new GAUSS for Windows, and provides a seemless Windows GUI for both GAUSS and GAUSSX. 

Recent additions include E-GARCH, simulated annealing, TABULATE, and PDFs, CDFs, inverse CDFs and random draws from 14 distributions. 

GAUSSX - Description

GAUSSX combines a full featured set of professional econometric routines, written in GAUSS, with a menu driven interface in one software package. The GAUSSX desktop, which runs under both DOS and Windows, is an intuitive interface that provides an excellent platform for both research and teaching, and can be used for running both GAUSS and GAUSSX. Alternatively, since the GAUSS source is included, econometric routines can be extracted and incorporated in stand-alone GAUSS programs.

GAUSSX runs from a desktop in which the user specifies the input and output files, and the directories for data. A command file is written using commands similar to SAS or TSP:

GAUSSX then executes this command file and the results are shown on the screen, and written to an output file, which is available for viewing after the end of execution. Thus GAUSSX replicates the edit/run/view cycle that seems to be most efficient in running econometric analysis. And since this cycle is menu driven, the learning curve is almost zero. One can run GAUSSX without knowing GAUSS at all. However, since any GAUSS statement can be used within GAUSSX, all the power of GAUSS is available. In addition, all the tools most commonly needed for econometric analysis are provided, at whatever level is required.

The current version of GAUSSX is 3.7

GAUSSX - Features

Linear Estimation Methods

Linear models include AR, ARCH, OLS, POISSON, QR, SURE, VAR, 2SLS, and 3SLS. Descriptive statistics, elasticities, automatic treatment of missing values, weighted analysis and White or Newey-West robust standard errors are standard. Lag specification such as y(-1) is supported, as are PDL structures. Diagnostics for single equations include Godfrey's test for residual serial correlation, Ramsey's RESET test for functional form, Jarque-Bera's test for normality of residuals, Breusch-Pagan test for heteroscedasticity, and Chow's test for stability. 

Non-linear Estimation Methods

Non-linear models include FIML, GMM, NLS , and ML. Step-size methods include BFGS, BHHH, DFP, GAUSS, NR, and SA (Simulated Annealing). Both the White and Newey-West robust estimators are supported. The defaults used during non-linear estimation can be altered heuristically during execution, or through a script file. Gradients, Hessian, and Jacobian are estimated numerically as the default; however they can be written as a procedure by the user. The maximum likelihood (ML) procedure permits the estimation of any specified likelihood - GAUSSX includes examples for non-linear ARCH, E_GARCH, GARCH, GARCH-M, MGARCH, MNL, MNP, NPE, POISSON, SUR, TOBIT, 2SLS and 3SLS. Coefficient restrictions can be imposed with PARAM, and investigated using ANALYZ, which can be used following either linear or non-linear estimation. Descriptive statistics, automatic treatment of missing values, and weighted analysis are standard. 

Time Series Analysis

A complete range of time series analysis is available under GAUSSX. ARIMA includes full identification, estimation and forecasting with graphical presentation. Systems of transfer functions can be specified, with a separate moving average structure for each equation. 

LDV Models

Linear LDV models include binomial probit, multinomial logit, and ordered logit and probit; in each case the marginal effects and elasticities, and their variances, evaluated at the mean, are available. For both probit and logit, Mills ratio is available allowing correction for selection bias. Heckman's two step procedure (HECKIT) incorporates Greene's covariance correction. Non linear multinomial logit and probit (MNL and MNP) are available using ML; for the latter, integration is carried out using the smooth recursive simulator for high dimension models. 

GARCH Models

A variety of ARCH/GARCH models are supported; these include linear ARCH, single equation non-linear ARCH, ARCH-M, GARCH, E-GARCH and GARCH-M. Multivariate GARCH (MGARCH) estimated over a system of equations, with the option of , weakly exogenous variables, is also supported, under both the VEC and BEKK formulation. MGARCH-M is also available. 

Exponential Smoothing

Methods include single, double, Holt-Winters, and seasonally additive or multiplicative Holt-Winters. Smoothing parameters can be user specified or optimally estimated by GAUSSX. 

Non-parametric Analysis

Non-parametric and semiparametric analysis under GAUSSX permits the estimation of the window width and the weights in the semiparametric index using cross validation under maximum likelihood. For the single index case, the FFT is used to speed calculation. Conditional response coefficients are determined for the density, conditional mean, discrete and smeared case. 

Neural Networks

The hidden and output weights in a feed forward network with a single hidden layer are estimated using non-linear optimization, rather than back propagation. Transfer functions include Arctan, Gaussian, Halfsine, Linear, Sigmoid, Step, and Tanh. Output processing includes levels, density, and maximum. 

Forecasting

Static and dynamic forecast values and residuals are available for all estimations. Systems of non-linear equations can be solved statically or dynamically. An impulse response function is available for VAR models. 

Kalman Filter

Analysis with the Kalman Filter allows for time varying transition matrices (ie. each element is a function), and the estimation of the elements of the Kalman matrices using ML. 

Data Handling and Conversion

Memory allocation and all file control is handled automatically. Data size for non-AR estimation is limited by disk capacity only. External data can be imported as delineated ASCII, packed ASCII, Lotus .WK* files, Excel .XLS files, GAUSS data files, GAUSS format files, and GAUSSX save files. Data can be exported as ASCII, GAUSSX, or GAUSS data files. Variables in a GAUSSX dataset can be user selected with the KEEP or DROP commands. 

Data Transformation

Data transformation (GENR) permits the use of all GAUSS operations and all the GAUSS functions, such as FFT, all GAUSS distributions, random number generators, etc. Thus all the power of GAUSS is available in GAUSSX. However sample selection (SMPL) makes coding far simpler, and data input/output is transparent. 

Descriptive Statistics

Each GAUSSX variable can have a descriptor (comment) associated with it. Data description includes means, standard deviations, minimum and maximum, sum, covariance and correlation matrices, autocorrelogram, partial autocorrelogram, and singular value decomposition (including variance decomposition). Divisia indices, seasonal adjustment, and principal components are supported. TABULATE (which replicates PROC TABULATE in SAS) tabulates data across two class variables by count, mean, SD, variance, minimum, maximum, sum and percent (row, column and total). This also permits standard frequency and crosstab. 

Econometric Tests

The TEST command includes the likelihood ratio test, the Chow test for stability, Goldfeld-Quandt test for homoscedasticity, F-test for linear restrictions, CUSUM and CUSUM-squared tests for stability, Granger's causality test, Dickey-Fuller test for unit roots, Engle-Granger and Johansen tests for cointegration, Thiel's decomposition of two vectors, Hausman's specification test, and Davidson and MacKinnon's J-Test for non-nested estimations. 

Simulation

Monte-Carlo simulation can be carried out over a block of code, using both bootstrap and jackknife methods. Output for the selected variables is shown dynamically on the screen, and final output includes cumulants and quantiles. 

Graphics

Graphical output (PLOT, GRAPH, COVA) is available either in text mode or in graphic mode, in mono or color. GAUSSX has full support for all GAUSS PQG routines. Line and scatter graphs/plots are supported. Printed output is available in both text and graphic mode. 

Distributions

A set of procs for evaluating density functions is included, which can be used from GAUSS or GAUSSX; these provide the PDF, the CDF, the inverse CDF and random sampling from the beta, binomial, central and non central chi-squared, exponential, central and non-central F, gamma, geometric, hypergeometric, log normal, negative binomial, normal, poisson, central and non-central t, uniform and weibull distributions. CDFMVN and sampling from a truncated multivariate normal distribution are also available. 

Programming Features

All GAUSS commands, logical goto, DO loops, and GAUSS procs can be used within a GAUSSX file. In addition, GAUSSX provides a number of programming commands; these include macro definitions for formulae, LOOP control for multisectored data, GROUP control (like BY in SAS) and recursive LIST names. 

Mixing GAUSS and GAUSSX

GAUSS statements can be included within the command file. GAUSSX variables can be made global (FETCH), and global variables can be stored in the GAUSSX workspace (STORE). Thus maximum flexibility is achieved by being able to mix GAUSS and GAUSSX commands. User written procs can be included within GAUSSX formula definitions. In addition, most GAUSS application modules can be run directly from a GAUSSX file. 

Extending GAUSSX

The complete source code, written in GAUSS, is included. Thus even if you don't want all the features of GAUSSX, you can extract a particular procedure and use it in your own GAUSS programs -- procedures such as inverse cumulative normal density function, Gibbs sampling, smooth recursive simulator (GHK), multivariate normal rectangle probabilities for any dimension, random sampling from a multivariate truncated distribution, and more. And because of its modular design, you can also add your own procedures to GAUSSX, or modify any GAUSSX procedure to fit your requirements. 

Desktop

The desktop is used to create and edit programs, run programs, view and print output, and manage data and projects. GAUSSX comes with a desktop for both DOS and Windows. Both supports SAA pull down menus, mouse and keyboard support, color support (for both the desktop and GAUSS), screen and printer control, user specified editor, viewer and utilities, as well as a wide range of tools. Project management is provided with up to 25 separate applications, each associated with different file names and paths. GAUSSX is network compatible - thus on a network, each client has its own project and configuration file. The desktop can also be used independently of GAUSSX to run pure GAUSS applications. 

GXEDIT - a Windows Editor

The editor that comes with GAUSSX for Windows is a full featured Windows editor, that can be used independently of GAUSSX (and GAUSS). Features include:

Help Facilities

During execution of a command file, pop-up help is available to explain the current screen, using Alt-H. At the GAUSSX desktop, help describes each of the desktop's functions. Under Windows, context sensitive help is available to provide the complete syntax of each GAUSSX command. 

System Requirements

GAUSSX can be run on a single machine, or on a network, under Windows or UNIX.

GAUSSX for WINDOWS runs under Windows 95, 98 and Windows NT. GAUSSX for Windows requires the new GAUSS for Windows 3.2.19 or higher, and about 2 MB of hard drive.

GAUSSX for UNIX is now available. It runs in both X-Window and Vanilla (or Terminal) mode. Networking is built in, so that individuals will each have their own configuration file. The econometric specifications for the UNIX version is identical to the DOS/Windows version. While we have not tested it on each UNIX platform, we have designed GAUSSX for UNIX to be machine independent by writing the entire package in GAUSS. Thus, if your UNIX machine runs GAUSS, it should run GAUSSX. GAUSSX requires GAUSS for UNIX 3.2.18 or higher, and about 1MB of hard drive. 

Send EMailScientific Web HomepageInternetadressliste HomepageMy Homepage
Stefan Steinhaus, webmaster@steinhaus-net.de