Software bugs are a fact of life, although we make great efforts to minimize them. By making a list of known bugs available, we hope to limit the inconvenience these problems may cause. Please contact us if you have any questions.
This page lists the significant known bugs that appeared in various Version 6 and Version 7 releases. Unresolved issues are listed first, followed by bugs that have been fixed in incremental updates. WinRATS users can download minor updates for certain versions at no charge from our website, in the form of "patch" files that update your existing application files. See the Software Updates page for details. See the RATS Updates page for details on ordering updates on CD.
May not work properly inside a procedure.
The ROW=FIND option on REPORT does not work properly. To work around this, you may be able to use DO loops and IF statements to check conditions and then generate a REPORT entry at a specific row as needed.
Version 7.2 includes fixes for the following issues:
The .NOT. operator had a higher precedence (one level above the == operator) than is indicated in Appendix A of the Reference Manual. Fixed in 7.2.
Scrolling down with the scrollbar "thumb" and then hitting PgUp would crash RATS. True in 7.0, possibly earlier.
Version 7.1 includes fixes for a couple of issues present in 7.0
The DAMP option on regression instructions was ignored unless you explicitly used the option LWINDOW=DAMPED (along with ROBUST and LAGS) to specify a damped lag window. This was fixed in 7.1 so that you could use the DAMP option without having to also specify LWINDOW=DAMPED.
The COLUMN option on IMPULSE did not work correctly. RATS would produce separate sets of response for shocks to all equations, not just the specified column. Fixed in Version 7.1.
ESTIMATE could crash if you used the MODEL option to estimate a model that was created using the GROUP instruction (rather than by a SYSTEM instruction). Fixed in Version 7.1.
FORECAST could cause RATS to crash if you used the ERRORS and STATIC option but did not also use the RESULTS option to save the forecasts. Fixed in 7.1.
Fixed an issue that could cause RATS to crash after scrolling through text using the vertical scrollbar "thumb" and then hitting the "PgUp" key.
The %BETASYS variable, intended to contain the coefficients of the VAR model stacked into a VECTOR, was not properly set by ESTIMATE. Fixed in Version 7.1.
Bar graphs that included one of the "hatch" patterns would not export to EPS (encapsulated PostScript) format files properly. Fixed in Version 7.1.
Improved handling of ODBC/SQL data handling. Eliminated a problem with reading the last observation in some data sets. ODBC sessions now closed automatically after reading, so multiple DATA commands don't produce multiple open ODBC sessions. Note: ODBC support is only available in the Professional version of RATS.
In some cases, Version 6.3 would fail to properly parse the %() function. Fixed in 6.35.
The %LOGCDF() function was never fully implemented in 6.2 and 6.3. Fixed in 6.35.
The SIGMA option did not work in conjunction with the FRMLVECTOR option. Fixed in 6.35.
In a user-defined FUNCTION, calling a matrix function more than once in an expression could produce incorrect results. Fixed in 6.35.
The %MQFORM function produces nonsensical results if the A matrix is RECTANGULAR and the B matrix is SYMMETRIC. Other combinations work properly. Fixed in 6.30.
Redirecting COPY output to the Output window could cause a crash with some formats. Fixed in 6.30.
For some Excel files, the SHEET option does not work, in which case only the first sheet in the file can be read (by omitting the SHEET option). Fixed in 6.30.
CVMODEL did not accept arguments for both the "A formula" and "B formula" parameters at the same time (worked fine if you only used one or the other). Fixed in 6.30.
Fixed an issue with using HISTORY and identity equations when using supplementary cards to list equations rather than the MODEL option. Fixed in 6.30.
Made a change to SUR to correctly report the degrees of freedom for each equation. Fixed in 6.30.
An extra comma was being included after an NA for series that started with a missing value in the first entry. Fixed in 6.30.
Made a change to PFORM so that it would accept series that contained only missing values. The corresponding entries in the stacked panel series are simply left as NA's (missing values), rather than returning an error message. Fixed in 6.30.
Regressions were not defining %NDFQ if the NOPRINT option was used. Fixed in 6.30.
%FRACTILES and %WFRACTILES did not return correct results if the input array included missing values. Fixed in 6.30.
Fixed an issue with printing reports generated with REPORT using decimal alignments. Fixed in 6.30.
The SMPL option on the SUR command is not working properly. It can produce crashes, or spurious "## REG13. Singular Regressions" errors.
Also, the RESIDS and COEFFS options that were added to SUR in Version 6.02 have not actually been functional.
Change to better compute means with ill-conditioned data.
Fix input check of VINDIV and VTIME to allow >=0, not just >0.
Added check for perfect fit, to avoid possible crash resulting from a division by zero.
Corrected Cauchy bound—previously this could fail to find roots in some cases
Fix bug that affects NOZUDEP option
Fixed a problem with the VERBOSE option when used with RATS files with a comment containing a %% character.
Fix a problem reading OLE workbooks from Excel 2000 or later. Most files would work properly, but some files would cause RATS to crash or fail to read in all data.
Fix problem with initializing MAX or MIN. Failed if max or min was 1.0
Fix to handling of internal pointers for asymmetry components in multivariate models .
Fix so that it returns NA if variance<=0 for univariate cases.
Fix so that it returns NA if variance<=0 for univariate cases.
When using the option PMETHOD=GENETIC in a MAXIMIZE estimation, the primary estimation method would sometimes start from the initial parameter values, rather than picking up the values returned by the genetic method preliminary iterations as it should (i.e. the initial iterations may effectively be ignored). Fixed in 6.10.
HISTORY did not work properly (could cause a program crash) when applied to a system that included identity equations. Fixed in 6.10.
The NOCENTER option on CROSS had no effect, so the series was always mean-centered even if you specified NOCENTER. Fixed in 6.10.
The function did not work properly, produced "junk" as a result, and could cause a crash. Fixed in 6.10.
Did not work properly when given a negative-valued argument (returned a negative value). Fixed in 6.10.
Could cause a memory overwrite, leading to a crash. Could also fail to find all roots in high degree polynomials. Fixed in 6.10
Version 6.02b is a minor revision made to WinRATS, WinRATS Pro, and UNIX RATS to fix a couple of issues discovered since the release of 6.02. For WinRATS users, a free downloadable patch update is available to update to 6.02b. UNIX users can contact Estima to receive an update via email. Note: MacRATS 6.02 was released later and includes these fixes: