Known Bugs

Known Problems in the Software

Q: In very large data sets, some of the procedures do not work and give the a message saying that some categories have fewer than 1 observations

A: This bug exists in version Beta 0.04.1. It is fixed in 0.04.2 which was placed on the web on July 18, 2002.

Q: The cancel button on the extract dialog does not seem to work.

A: This was a bug in Beta Version 0.04.1. It was fixed in 0.04.2 placed on the web on July 18, 2002.

Q: Problem importing from SPSS version 11

A: This is fixed in beta version 0.04.3, dated July 29, 2002. It required a bit of a work-around, so please report and residual problems after downloading the current version.

Q: The replication procedures seem to add an extra weight variable when you use the "copy model to modify and re-run" option from the completed run queue.

A: This bug has been fixed in beta version 0.05.0 available on the website on January 7

Q: The procedure to dummy code variables does not treat missing values properly.

A: This bug seems to have been introduced in version 5, which was released in January, 2003. A fix is available on the website as of March 5, 2003. The fixed version 5.12. The patch currently on the website fixes this and all previously reported bugs. Users are strongly encouraged to install the patch.

Q: A user reports a crash when importing data from SPSS

A: This problem stems from a bug in the program. We have located and repaired the bug, and will post the fix in Version 0.05.14 by June 3, 2003 to the website.

Q: Missing values don't import properly from SPSS

A: This will be fixed in the next release. Eliminate missing value RANGES (unique values are ok) from SPSS files. Use AM to recode these to missing. To use AM to recode values to missing, right-click on the variable in the Variable List and select "Set Missing Values" and proceed from there

Q: In linear regression using replication, p-values are incorrect.

A: This was a bug in Version Beta 0.05.14. It affected only the reported p-value for this procedure. It is fixed in Version 0.05.15, which is available on the website as of August, 20, 2003

Q: AM doesn't allow opening *.sAM model specification files for replication procedures saved previously.

A: This is a bug. We have corrected it and will incorporate the fix in Beta version 0.05.15, which we will release before September 1.

Q: In v05.15 p-values for linear regression with replication are still not correct.

A: Thank you. We caught this production problem, and are preparing a re-release before the end of the day. Sorry for the inconvenience. The new release will be version 0.05.16.

Q: When a data set has only one PSU in a stratum the taylor-series procedures that are not maximum likelihood procedures do not capture the variance of the affected clusters. In this case, standard errors are underestimates.

A: This is a bug through version 0.05.16. It will be fixed in version 0.06.00, which will be released in late October or early November, 2003

Q: The t-statistic associated with constant terms in the multinomial logit model are incorrect in Version 0.05.16 and earlier.

A: This is a reporting problem that does not affect parameter estimates, standard error estimates, or t-statistics for other parameters. It will be fixed in version 0.06.00, which is scheduled for release at the end of October, 2003.

Q: The release of version 0.06.00 had a bug that could lead to inaccurate standard errors for the taylor-series frequencies and descriptive procedures under some sample designs. The discrepancies were typically very small.

A: This bug was fixed as soon as it was detected. The fix appeared in Version 0.06.01, which was available on the web as of October, 27, 2003.

Q: The general export procedure outputs missing values as a very large number, rather than a system missing value.

A: This bug has been fixed in the version of the AMDFTransfer component placed on the web on 12/01/03. We made a corresponding patch available at the same time. Current users who have the transfer component installed can just install the patch.

Q: The plausible values procedures do not accurately report the filter that was set when the statistics were calculated, but the statistics are correct.

A: This is a reporting bug. Users should be aware of it. it will be fixed in a subsequent release.

Q: In beta versions 6.00 and 6.01 (fixed in 6.02)A programming bug can cause a crash or system instability in some circumstances, most common in large, simple random samples.

A: This bug is fixed in beta release 0.06.02, placed on the website on April 15, 2004 (tax day). Registed users can install a patch. New users will download the new version, and will not require the patch.

Q: When cancelling a run, AM crashes.

A: This is a bug in the current release, and will be fixed in the next release. The procedure tried to create a table with over 1000 rows (the independent variable has over 1000 unique values). In the process, the system tries to create a 2000x2000 matrix. The crash results as a function of a page fault or exhausted virtual memory, and can occur whether or not you try to cancel the run. Prior to the next release, try to avoid tables with so many rows.