JML analyses no longer working

[njsbrown]

JML analyses that worked successfully a few months ago are now causing ConQuest 3.0.1 to crash.

This is true in two different licensed copies, and a newly downloaded trial version, on several different computers, using identical command files and data files that worked in the past.

ConQuest crashes with no error message, causing a Windows error message: "Conquest Application has stoped working: A problem caused the program to stop working correctly. Windows will close the program and notify you if a solution is available."

The crash occurs at different times, usually during estimation, often immediately after sorting cases, but occasionally after estimation during a show command.

MML analyses are working normally.

This seems like a strange error, and I would have assumed it was computer-specific, but we've gotten the same results on several computers and in several copies of ConQuest 3.0.1.

Any help would be greatly appreciated.

Best, -Nathaniel

[Eveline Gebhardt]

Could you please email us the command and data file so we can debug (gebhardt@acer.edu.au)?

Eveline

[castleca]

Hello,

I am having an identical issue when I try to run JML analyses using Conquest V4. Was this problem ever resolved, and if so, how may I address it?

I have attached syntax for a minimum working example, using the dataset from Tutorial 2.

Thank you for your attention,
Courtney

[Alvin Vista]

Hi Courtney,
I was able to run your syntax, although there was a bug in requesting the show file. The itanal and plot outputs are fine.
We will inform you as soon as the bug in the show command for JML is resolved.
Cheers,
Alvin

[Alvin Vista]

Hi Courtney,
I think the bug was that the default standard error type for JML estimation is not activated. The default for standard error type is empirical but JML can only do quick standard errors.

A workaround for this is to specify the type in the estimate command option.
estimate !method=jml, stderr=quick;

I hope this works for you.
Cheers,
Alvin

[castleca]

Hi Alvin,

Thanks for your prompt reply. I changed the type of standard error, but the analysis still crashed. I attached a screenshot of Conquest at the time of the crash, in case it is useful. Windows throws the error - not Conquest.

Any additional help would be much appreciated.

Thank you,
Courtney

[Alvin Vista]

Hi Courtney,
This is a bug and we're working to fix this. It seems that the JML routine is having problems writing the show file output.
A possible workaround, while waiting for the fix, is to use another method in a preliminary estimation, then reestimate using JML before requesting for the show file.

estimate !method=gauss, stderr=none;
estimate !method=jml, stderr=quick;

Cheers,
Alvin

[castleca]

Hi Alvin,

No luck - same error message after the second estimation, and no .shw file produced. Thanks for all of your help - I'll anxiously await news of the bug fix. 

Courtney