See RATS 7.0 for an overview of Version 7. This page provides detailed information on the new features, options, and functions introduced in Version 7 and in earlier incremental updates released since 6.0. Most of this information is taken from the first section of the Version 7 User's Guide.
RATS 7 greatly expands the number of instructions covered by “wizards.” (There are so many now that there are three menus covering them: Wizards, Data (which covers data reading and transformation) and Statistics. The wizards generate (and execute) RATS instructions, complete with options, parameters, and supplementary cards, which makes this feature particularly useful for new users who are learning the RATS language. For more experienced users, the most helpful wizards are probably those for reading data (Data (RATS Format) and Data (Other Formats) on the Data menu) and those for functions and variables (Function and RATS Variables on the Wizards menu).
The method of obtaining “regressor lists” in the new and existing wizards has been simplified so you can either type or edit the list directly, or you can use a specialized popup dialog box which allows variables to be selected out of a scrolling list.
For version 7, we have made major changes to standardize the format of instructions and the names of options. With few exceptions, instruction parameters are now used only for input or output series and entry ranges. Other types of information (whether input or output) are provided using options. Options have names and can be input in any order, making them easier to remember, use, and read than parameters, which need to be in a very specific order. While we still support the older syntax (RATS is largely backwards compatible to Version 4), in most cases we haven’t documented in the main manual the clumsier ways of writing instructions. There’s a separate pdf file that includes the older instructions and options.
Among the changes are new FROM, TO and STEPS options for the forecasting instructions, replacing their use of parameters; and CV and CVOUT options for inputting and retrieving a covariance matrix in a wide range of instructions, replacing options (ISIGMA on SUR and NLSYSTEM) and parameters (the 3rd parameter on IMPULSE, ERRORS and SIMULATE).
| DSGE | Takes a model with expectational terms (indicated by leads) and solves out for a linear or linearized state space model. |
| SWEEP | Handles a wide range of analyses based upon a regression of a set of targets on a common set of variables, whether it’s extraction of residuals from components or heterogeneous regressions. |
| QZ | Does a generalized Schur decomposition of a pair of matrices. This is a step which is now included within the DSGE instruction. |
There have been major changes to the way that graphs are produced and displayed. There are now 30 “representations” for line types, fill types and symbols, describing the pattern or shape, color or gray level and thickness. Each representation comes in a color form and a black and white form. While the color form is almost always the easiest to use on the screen, most hard-copy output still needs to be in black and white. You can now easily switch between the two views using a toggle button on the toolbar when you have a graph displayed. You can also define your own combination of attributes for a representation.
The graphing instructions also have new FOOTER and FRAME options, and almost any text can be shown in multiple line fields by using \\ in a string to signal a line break.
The four “least squares” instructions (LINREG and NLLS for single equations, SUR and NLSYSTEM for systems) now have a common set of options for handling gmm estimation in a flexible way. The UPDATE option, for instance, allows for several methods of updating weight matrices which were available in the past for some of the instructions but not others. The new JROBUST option allows for the J specification statistic to be “robustified” by either recomputing the statistic or using a non-standard distribution. The biggest change here is to SUR, which did not have any options for GMM.
The DLM instruction was almost completely reworked from version 6. Although older code using this will still work, the input into this can now be much more natural, and there are many more options for extracting information.
The CLUSTER option allows for clustered standard error calculations in any instruction which allowed for robust standard errors before.
Estimation instructions which use PARMSETS now have REJECT options which allow the function to return an NA (missing value) immediately when parameters go out of range.
Almost ninety new functions have been introduced since 6.0. The principal changes are the addition of a number of new random draw functions, particularly “convenience functions” for drawing from standard posteriors; density, distribution and inverse distribution functions not previously available, and new functions which access the internal “perpetual calendar.”
There are several new operators which can be applied to matrices. A+r and A-r for array A and real number r are interpreted to add or subtract r from all elements of A. Matrix exponentiation (repeated multiplies) can be done using A^n (or A**n) where n is an integer. Elementwise exponentiation can be done with A.^r, which will take each element of A to the r (real) power. A~B, A~~B and A~\B do horizontal, vertical and diagonal concatenation of matrices A and B (largely replacing the %BLOCKGLUE and %BLOCKDIAG functions).
| AR1 | The new PRHOS option allows retrieval of the individual AR(1) coefficients when doing heterogeneous panels. |
| ASSOCIATE | New %EQNSET… functions make this largely obsolete. |
| BOOT | The options BLOCK, METHOD and GEOMETRIC allow several types of block bootstrapping. |
| BOXJENK | Offers the choice of least squares (as before) and maximum likelihood with the MAXL option. This now has PMETHOD and SMPL options. |
| CALENDAR | This supports a new and easier to use syntax. |
| CATALOG | Adds the LIKE option for restricting the listing. |
| CMOM | Adds a MODEL option to pull the variables in the cross moment matrix out of an existing equation. |
| CORRELATE | Has an ORG option to allow choice between two layouts for the output. |
| CROSS | Similar change to CORRELATE. FROM and TO options were added to replace parameters. |
| CVMODEL | The V option allows direct modeling of the covariance matrix rather than just its factors. Adds the REJECT option. |
| DATA | Adds the SHEET option for reading from spreadsheets. sql option and FORMAT=ODBC were added to the Professional version to read from sql databases. |
| DBOX | Adds options for editing matrices and selecting elements within a fixed matrix. Added SPINBUTTON object type. |
| DDV | The COEFFS option allows you to retrieve the full coefficient matrix for multinomial logit. The EQUATION option can be used to feed in the equation to be estimated. |
| DECLARE | Can be included at any location within a procedure (not just at the start). |
| DIFFERENCE | Adds options D and PADDING for fractional differencing. |
| DLM | This had a major reworking. TYPE=SIMULATE and TYPE=CSIMULATE were added for random simulation and conditional simulation. There are options to retrieve many more intermediate calculations. Support was added for partial observables in multivariate models. The input to it through the options has been made more flexible and natural. There are new options for an inverse chi-squared prior on observation equation variance. The new PRESAMPLE option allows DLM to deal automatically with models which are partially stationary and partially non-stationary. |
| EIGEN | The GENERALIZED option allows it to solve the generalized eigenvalue problem for symmetric matrices. |
| END | This now needs a RESET option to reset memory. This prevents accidental reset. |
| ERRORS | This adds the STDERRS option for retrieving the calculated standard errors of forecast. |
| ESMOOTH | The option FITTED allows you to obtain (easily) the one-step forecasts. The CONSTRAIN option constrains estimated parameters to the unit interval. |
| FILTER | This adds TYPE=SPENCER for Spencer moving averages, WEIGHTS for general filters, and the EXTEND option for methods of extending the filter at the ends of the data sets (other than truncation of the output). The REMOVE option allows a convenient way to detrend or deseasonalize data by regression. |
| FIND | Adds the REJECT option. |
| FORECAST | This adds the STDERRS option to get the standard errors of forecast without a separate ERRORS instruction. The options STATIC and ERRORS allow static (in-sample) forecasting without using a separate STEPS instruction. |
| GARCH | This adds the I option for igarch. The new VARIANCES option controls the variance model for individual series in a CC or DCC. MV=VECH and MV=EWMA options for multivariate garch have been added. The MVHSERIES allows the variance series to be retrieved in a form usable for multivariate garch-m models. The DERIVES option returns the derivatives of the likelihood with respect to the parameters. |
| GCONTOUR | This adds FRAME and FOOTER options. |
| GRAPH | The adds the FOOTER and FRAME options. There are new FAN, DOT and SPIKE options for STYLE and OVERLAY. The OVRANGE options allows the left and right scales in an overlay graph to have separate vertical ranges. It adds the standard SMPL option. |
| GRPARM | This adds the PATTERNS option for globally choosing to use black and white patterns rather than colors. The IMPORT option pulls in user-definable choices for representations of graphs. The RECALL option, when combined with the %GRPARM() function allows you to save and restore settings. |
| GRTEXT | This has new choices for the POSITION option: BOTTOMMARGIN, LEFTMARGIN and TOPMARGIN. |
| HISTORY | This adds FROM, TO and STEPS options. |
| IMPULSE | This adds the STEPS option. The CV and COLUMN options are designed to replace parameters. |
| INCLUDE | The new COMMENTS option allows comments to be calculated strings, not just input text. INQUIRE The EQUATION option allows analysis of the range determined by the variables in an equation. The VALID option returns a dummy variable with the valid data points in an equation or regression. |
| KALMAN | The options X, Y and V have been added to do dummy observation priors. |
| LDV | Adds the EQUATION option to specify the equation being estimated. Adds the CLUSTER option. |
| LINREG | This adds CENTER, JROBUST, UPDATE, ZUDEP and ZUMEAN options for more general gmm. |
| LOCAL | Can accept dimension fields and can be placed anywhere in a procedure or function. |
| LQPROG | This has options for all inputs. The GE option allows for constraints that are naturally stated as ">=" constraints. NNEG (for quadratic programs) allows parameters to be not constrained to be non-negative. |
| MAKE | The PANEL option allows creation of a matrix from a panel data series. |
| MAXIMIZE | Adds the REJECT and CLUSTER options. |
| MCOV | Adds the MODEL option to take variables out of an existing model. It also adds the CLUSTER and ZUMEAN options. |
| MVFRACTILE | Has been combined into MVSTATS. |
| MVSTATS | Includes options formerly in MVFRACTILE. It adds an option for handling end-of-sample data windows. |
| NLLS | This adds SV and SPREAD options for weighted least squares either including estimated parameters (SV) or not (SPREAD). JACOBIAN allows for Jacobian terms for maximum likelihood. DERIVES returns the derivatives of the residual with respect to the parameters. |
| NLSYSTEM | This adds the FRMLVECT option for a model which generates the residuals in a VECTOR form, the MODEL option for more convenient input. JACOBIAN allows for Jacobian terms for full-information maximum likelihood. DERIVES returns the derivatives of the residuals with respect to the parameters. It adds REJECT and CLUSTER and the full set of gmm options: CENTER, JROBUST, UPDATE, ZUDEP and ZUMEAN. |
| NONLIN | PARMSETS can now include VECT[VECT] and other more complicated array structures. |
| PREGRESS | This adds METHOD=BETWEEN. It performs an automatic Hausman test with METHOD=RANDOM. INDIV and TIME options are available to retrieve components. |
| This adds the standard SMPL option. | |
| PRJ | This has LOWER and UPPER options for general truncations. |
| PRTDATA | Adds the LIKE option for restricting the series printed. |
| REPORT | This adds the options UNIT and FORMAT for direct export of the table. New option choices are ACTION=SORT with BYCOLUMN and SPECIAL=BRACKETS. |
| RLS | Adds the CONDITION option to condition on more than the minimal number of observations. |
| SCATTER | Adds STYLE=SPIKE and STYLE=STEP. Adds FOOTER and FRAME. And now has the standard SMPL option. |
| SET | Now has a SMPL option to leave observations out of the calculation. |
| SIMULATE | This adds FROM, TO, STEPS and CV options. |
| SPGRAPH | SPGRAPHs can be nested so you can add special text (with GRTEXT) to graphs within a matrix. It also adds the FOOTER option. |
| STEPS | This has FROM, TO and STEPS options. Note though that it is largely obsolete due to new options on FORECAST. |
| STORE | Adds the SHEET option for reading from spreadsheets. SQL option and FORMAT=ODBC were added to the Professional version to read from SQL databases. This has the LIKE option for restricting the set of variable names imported. REFRESH imports only the data for series already on the file. |
| STWISE | Adds options for HAC standard error calculations (ROBUSTERRORS, LAGS, etc.) |
| SUMMARIZE | Accepts non-linear expressions (on the command line). It does computes the (approximate) variance using the delta method. |
| SUR | This adds the MODEL option, and a large number of options for full support for gmm. |
| TABLE | Includes TITLE and PICTURE options |
| THEIL | Adds the FROM, TO and STEPS options. |
| UFORECAST | Now includes BOOTSTRAP and SIMULATE options for simple forms of bootstrapping and random simulation. |
| VCV | Adds the WINDOW option. |