TFQA: Tools for Quantitative Archaeology
     kintigh@tfqa.com   +1 (505) 395-7979

TFQA Home
TFQA Documentation
TFQA Orders
Kintigh (ASU Directory)

Program Problems and Fixes

CNTCNV Column Totals  Computed Incorrectly When Data Are Aggegated.  (Fixed in V6.2  on 9/27/2022)  Until this version, CNTCNV was reporting column totals for an observation based on the total sample for that observation, instead of only the row-column combinations  that are included in the aggregation list.  Counts for row-column combinations that are excluded by the aggregation should have been ignored. If no aggregation file was used, the results should be correct in earlier versions. 

Graphics Programs Can't Find the Input Files. This isn't actually a bug. KMPLT, DIVPLT, LDPLT, BAYES, SCAT, and FORD run in 16 bit Windows mode and can only handle file names with at most 8 character names (including letters and numbers but no special characters except _) and an optional period and 3 character extension. If you are running these programs, the easiest solution is to rename the input files so they meet this specification. You can also use the short (8.3) versions of the file names (see Wikipedia article), but it is easier to rename them. This issue is described in a little more detail in Running TFQA Under Windows.

Importing Graphics into Microsoft Office 2007. As described in Running TFQA Under Windows, Microsoft offered a filter for Microsoft Word 2003 that could be downloaded to allow HPGL plot files produced by the plotting programs (KMPLT, DIVPLT, LDPLT, BAYES, SCAT, and FORD) to be imported into Word and printed. This filter does not work for Word 2007 and no alternative has been offered. Viewcompanion Standard will covert the HPGL files to a variety of other formats including jpg and nicely prints the HPGL files. More information on printing graphics in ViewCompanion is provided in Running TFQA Under Windows.

Seeing Output from TFQA Graphics Programs. The few TFQA programs that create full screen graphics do run properly under Microsoft Windows Vista (KMPLT, DIVPLT, LDPLT, BAYES, SCAT, and FORD). The other programs, inclu.ding the companion computational programs (KMEANS, DIVERS, and LDEN) continue to work fine under Windows Vista and Later Versions The problem is caused by a lack of backward compatibility that for technical reasons has no easy fix. Fortunately there is a work-around-see http://tfqa.com/vistafix.html.

Dsplit runs case i as case iia. In versions of dsplit earler than 3.1 (8/16/2007), if the input data included non-zero calibration and sunspot errors they were erroneously included in the calculations for a case i analysis (i.e., the analysis done was actually a case iia analysis). In the earlier versions there was no problem with the other analysis cases. Also, if these two error terms were not included in the input data (or they were 0), the case i analysis was done correctly. The calibration and sunspot error terms are now treated correctly in all cases. Registered users can email me for the current version of dsplit.

Programs that Plot Graphics on the Screen - Top line of graphics screen is off the top of your monitor or the program seems to freeze with a blank screen. This seems to happen with some large monitors. This is fixed on current versions (7/22/07 and thereafter) of the toolkit. Users who have purchased an earlier version of the program can email me to get updated versions of Bayes, DivPlt, Ford, KMPlt, LDPlt, and Scat.

Programs fail to run under Windows XP, SP2 with the error: "16 bit Windows Subsystem - C:\WINDOWS\SYSTEM32\AUTOEXEC.NT. The system file is not suitable for running MS-DOS and Microsoft Windows applications. Choose 'Close' to terminate the application."

Evidently, the Microsoft Service Pack 2 update may sometimes delete the AUTOEXEC.NT file that is needed for the programs to run.  See that article or http://support.microsoft.com/?kbid=324767 for more information.

To verify whether you have the file,type "%windir%/system32/ in the address bar of a Windows Explorer window (not Internet Explorer). If there is no AUTOEXEC.NT file proceed as follows:

1.    Browse to "%windir%/repair/" (usually "C:\WINDOWS\repair

2.    Right-Click and Copy the >AUTOEXEC.NT file

3.    Browse to "%windir%/system32/" (usually "C:\WINDOWS\System32

4.    Right-Click inside the window and Paste the file

The error condition described here may also be the result of corruption of the AUTOEXEC.NT file, in which case the above procedure may be helpful to restore a valid file.

Corruption of the AUTOEXEC.NT file has also been related to virus programs. You may want to virus-check your system.

Kmeans/KMPlot (7/11/2005). When k-means is run with a request for a maximum number of clusters approaching the number of observations (technically, when the SSE of a given cluster solution goes below about 0.001 of the original SSE), the kmplt program may produce incorrect (and obviously so) SEE reduction plots for the random runs (e.g., where the SSE increases with more clusters). The original kmeans computations were correct, but kmeans provided incorrect information to kmplt. This problem has been fixed. If you have purchased the programs prior to 7/10/2005 and have run into this problem, email me for updated version of kmeans/kmeanscsv). The new version corrects this problem, in part by reducing the SSE improvement of a configuration that is considered significant. This may occasionally lead to longer run times but will achieve the same or better (lower SSE) results.

Can't Figure out how to Run the Programs under Windows. See instructions.

Programs Fail with Runtime 200 Message or Appear not to Run. When many of the programs in very old versions of Tools for Quantitative Archaeology are run on machines of with 200MHz or faster chips, the programs immediately fail with a Runtime Error 200 or appears not to run at all.  If you are not having this problem, you have nothing to worry about; no analytical results are affected and you need not read any farther. If you are having this problem, please read on.

This is due to a compiler error in the Borland Pascal Compiler that was used to compile all of the programs. I have now (1/2001) recompiled all of the programs so that this problem is avoided. The symptoms are that if you run the program by double-clicking the .exe file the program starts and ends so quickly that you see no real effect other than a window flashing open and closed. If you run the programs from an MS-DOS window, you get a message like "Runtime error 200 at 037C:0091." Purchase of the current update will eliminate the problem.  The recompiled programs are smaller and faster but will not run on extremely old machines (8088 or 8086 machines or 80286 and some 80386 machines lacking math co-processors.

Boone Program Z Score Calculation Error. Versions of the Boone Program below version 3.6 (prior to May 2000) are incorrectly calculating the Z-score related variant of Boone's measure. The other computations are OK. (Some even earlier versions of the program did not do this computation at all.) If you purchased an affected program, email me and I can will send an updated program as an attachment. It is my sense that this calculation is rarely used, but I apologize for any problems this may have caused

Page Last Updated: 27 September 2022