Statistical Features

Estimation Techniques
• Multiple regressions, including stepwise.
• Regressions with autoregressive errors.
• Regressions with heteroscedasticity correction.
• Seemingly unrelated regressions and three-stage least squares.
• Non-linear least squares.
• Two-stage least squares for linear, non-linear, and autocorrelated models.
• Maximum likelihood estimation, supporting a wide variety of problems including ARCH, GARCH and related models. Supports multivariate likelihood functions, and offers a simplex estimation method for non-differentiable functions.
• Non-linear systems estimation.
• Generalized Method of Moments.
• Constrained optimization (Version 5.0).
• Built-in hypothesis testing instructions.
• Logit and probit.
• Fixed/random effects estimators (Version 5.0).
• Consistent covariance matrices, including "White" and "Newey-West".
• State-space models (Version 5.0).
• Neural Network models.
• Linear/Quadratic programming.
• Kernel density estimation (Version 5.0).

Time Series Procedures
• ARIMA models including multiplicative seasonal models (supports arbitrary lag structures).
• Vector autoregressions.
• Impulse responses.
• Variance decompositions.
• Structural VAR's (Version 5.0).
• Error Correction Models (Version 5.0).
• Kalman filter for sequential estimation.
• Transfer functions.
• Intervention models.
• Spectral analysis.

Forecasting
• Time series models.
• Regression models.
• Exponential smoothing.
• Simultaneous equation models (supports unlimited number of equations).
• Simulations with random or user-supplied shocks.
• Forecast performance statistics.

Graphics
• Line, symbol, bar, stacked bar and filled graphs for time series.
• Dot, symbol, line, bar and filled X-Y graphs.
• Dual-scale (overlay) time series graphs and (Version 5.0) X-Y graphs.
• Arrays of graphs on a page.
• Log scales (Version 5.0).
• Exports graphs to PostScript, HPGL, PIC, Windows Metafile, and Macintosh PICT.
• Export can be done automatically, as graphs are produced (Version 5.0).

Data Entry
• Reads and writes ASCII, DIF, PRN, DBF and binary files.
• Reads and writes Lotus Worksheet (including WK3) and Excel XLS spreadsheet files.
• Write HTML tables (Version 5.0).
• Supports our own RATS data file format, which is fast and easy, supports all frequencies, and allows mixing of frequencies on a single file.
• RATS format now portable across DOS, Windows, Macintosh, and UNIX platforms.
• On-screen data editor.
• Menu-driven RATSDATA utility program for maintaining, graphing, importing, and exporting data.
• Can handle virtually any data frequency, including daily, weekly, intra-day, and panel data.
• Easily converts among different data frequencies.

Data Transformations
• Flexible transformations with algebraic formulas.
• Easy to create trend series, seasonal, and time period dummy variables.
• Specialized differencing and filter operations.

New Instructions

RATS Version 5 offers five new instructions:

• CVMODEL estimates structural VAR’s.
• ECT adds error correction effects to a VAR.
• DLM works with state space models, employing Kalman filtering and smoothing.
• PFORM creates panel series from other sources.
• PREGRESS performs fixed and random effects estimation.
• DENSITY computes empirical density functions.

Non-Linear Estimation
• The ability to use arrays as non-linear parameters.
• A new PARMSET variable type that makes it easy to combine several sets of non-linear parameters, allowing models to be created in pieces and combined only when estimated.
• The ability to impose equality and inequality constraints.
• A new "genetic" estimation method which allows a much broader search of the parameter space.
• The ability to define vectors of formulas, including the option of creating them inside a loop.

Vector Autoregressions and Systems of Equations
In addition to the new CVMODEL and ECT instructions, several "convenience" features have been added. For example, it is now much easier to save VAR residuals, forecasts, and impulses responses, using the new "RESULTS" option. Also, the SYSTEM instruction now includes a MODEL option that allows you to define a VAR as a model, which makes setting up instructions like IMPULSE and FORECAST much easier.

Probability and Random Numbers
New functions have been added for computing probability distributions and their inverses, as well as factorials and binomial coefficients. Additional documentation covers techniques like Gibbs sampling and approximate randomization in detail.

Revised Documentation
The RATS documentation has been reformatted into two standard 7"x9" books, and will be included in electronic form (Adobe PDF files) with each copy of RATS. The manual features more examples, more technical details, and offers expanded/new coverage of many topics, including Structural (identified, Bernanke-Sims) VAR’s, Blanchard-Quah decomposition, ARCH and GARCH models, hazard models, state-space models, simulations and bootstrapping, and non-linear estimation.