RATS has the right mix of sophisticated built-in statistical functions (for instance, for state-space modeling,
GARCH estimation, ARIMA models and vector autoregressions), menu-driven operations and programming features to
help you get your tasks done quickly *and correctly*.
See RATS Features for a complete list.

In addition, we have a wide collection of resources available to help you get the most out of your software. This includes over a thousand worked examples from more than twenty major textbooks, replications of over fifty important papers on a wide range of topics, and hundreds of procedures to extend the standard "built-in" calculations.

RATS has been heavily tested. See NIST results for a look at how we do on standard benchmarks. In addition, all the examples described in the previous paragraph were checked carefully for accuracy. In the vast majority of cases, if there was a problem found replicating a result, it was an error in the original work, not with the RATS calculation.

Much of the current empirical work in econometrics uses "math packages". These are designed to work primarily with data in "matrix" form, and they do built-in matrix functions very well. However, particularly in time series work, this view of data doesn't work well, because the data matrix for one operation with one set of lags is different from the one needed with a different set. If you look at a program for, for instance, the estimation of a Vector Autoregression, very little of it will be the matrix calculations to compute the coefficients—most will be moving information around to create the required matrices. This can be a very error-prone process and, in fact, there have been quite a few published papers which have had serious errors in data handling.

By contrast, RATS keeps the data in a fixed location, anchored to its time periods, and automatically adjusts the sample range to allow for lags of the explanatory variables and for any other type of missing values.

RATS also has carefully tested routines to estimate things like GARCH or state-space models which would have to be written in matrix language. You might be able to find or buy subprograms to handle those, but the quality is rather uneven—some are very good, some aren't.

Because RATS is specially designed for time series work, it automatically takes care of adjusting the sample range to allow for lags. You don't have to create separate variables for lags—you just tell RATS to use lags of what you have.

The RATS GARCH, DLM (for state-space models) and BOXJENK instructions have features not found in any other comparable software. The programming language has everything you would expect: scalars, matrices, procedures and functions, but also includes specialized "data types" which simplify tasks which could be very complicated if they didn't exist. For instance, you can "add" non-linear parameter sets, allowing you to combine sets for separate equations or separate parts of a model (mean and variance, for instance) when estimating a complete model. You can create complex structures from basic elements such as data series for easy of manipulation.

RATS also has user-definable menus and dialogs for very sophisticated types of programs. CATS is the best example of this—it is entirely written in the RATS programming language.