by TomDoan » Tue Sep 06, 2011 3:30 pm
Many of the improvements in this new version are due to our work with four new or updated textbooks. The new editions of Greene's Econometric Analysis and Wooldridge's Econometrics of Cross Section and Panel Data included quite a bit of additional coverage for techniques involving panel data and data with more general groupings, and Baltagi's Econometric Analysis of Panel Data included some methods which were not as easily done with RATS as we would have liked.
The principal changes made for the examples from those three books are to the PREGRESS and PANEL instructions. PANEL, PREGRESS (and PSTATS) all include a new GROUP option, which allows estimation of individual effects with arbitrary groupings that aren’t really “panel” data. PREGRESS has many other new options, allowing instrumental variables estimation with fixed or random effects, a VCOMP option to allow a wide variety of choices for estimating the component variances in a random effects model and ROBUST and CLUSTER options for computing clustered standard errors for fixed or random effects estimators.
PANEL adds EFFECTS, VRANDOM, VINDIV and VTIME options (similar to those already in PREGRESS) to compute standard panel data transformations for inputs to more complicated techniques (Hausman-Taylor for instance) that can’t be done directly with PREGRESS. With the GLS option, you can choose how those transformations are done. In particular, GLS=FORWARD uses only the “forward” means, so the transformed data at time t uses only data from time t on. This can be useful in dealing with the bias problem in dynamic regressions in small panels. The new ID and IDENTRIES options are used extensively in the textbooks examples to organize calculations in unbalanced data that require some form of integration or summing across time for individuals.
Interface Features
A new File menu operation lists “Recent Directories” used with RATS.
You can now create “scratch” windows, which are text windows that are both “input” and “output”. If you have a need a quick calculation or want to test an alternative regression without either the instructions or output getting inserted into your main work windows, the scratch window is the way to do that.
INFOBOX with a progress indicator will now estimate the time until completion if it will take more than a few seconds.
Edit-Find and Edit-Replace now have popup boxes with past find and replace strings.
Functions
%BLOCK(entry,blocksize) returns the block number of entry in dividing data into blocks of equal size.
%PANELSIZE() returns the number of individuals in the current CALENDAR with the PANEL option. (The already existing %PANELOBS() gives the number of time periods).
%LOGDIRICHLET(x,d) gives the log of the Dirichlet density for the VECTOR of probabilities x given VECTOR of shapes d.
%MODELLAGMATRIX(model,k) returns the N x N matrix of lag k of the dependent variables of a model.
%SLIKE(template) returns a vector of series handles matching a name template.
%SROW(S,row) returns a vector of series handles for a row out of a matrix of series.
%SCOL(S,col) returns a vector of series handles for a column out of a matrix of series.
Other instructions
BOOT
Now has a PANEL option to shuffle entire individual records in (balanced) panel data.
BOXJENK
Allows missing data with the MAXL option and differenced data.
DATA
The Pro versions of rats now include FORMAT=FRED, which pulls data directly off the St. Louis Federal Reserve’s FRED database into RATS if you have the series names.
DATA will now set the end-of-data automatically based upon full list of series read. This used to do this based upon the first series read, which might not be correct if you are reading data from different original sources.
ENTER
Adds the SEQUENCE option to allow use with LIST and CARDS.
GARCH
The new HADJUST and UADJUST options allow “on-the-fly” calculations using current variances or residuals. This simplifies the coding for many types of garch-m models and related models.
PFORM
Adds an INPUT option to allow combining series into a balanced panel when your input has separate series for each time period.
PRJ
Adds the COEFFS option to override coefficients from the estimated equation, and the RESIDS option to compute and save residuals.
PSTATS
Adds the GROUP option as described above for general groupings. Most of the statistics it computes are now accessible.
READ
8.1 implements the SHEET, TOP, LEFT, BOTTOM and RIGHT options that were already available on DATA. These allow you to isolate the data from other information. You can also read complicated array forms (such as VECTORS of VECTORS).
SET and SSTATS (CSET, GSET as well)
These add the PANEL option (actually the NOPANEL option) to suppress panel data handling of lags. This makes it easier to do panel bootstrapping.