[Bug]: Cross-platform inconsistencies in Model-based Clustering

[LucasMolleman]

JASP Version

0.95.1

Commit ID

No response

JASP Module

Machine Learning

What analysis are you seeing the problem on?

Model-Based Clustering

What OS are you seeing the problem on?

Windows 11

Bug Description

Hi JASP team,

Thanks for making this great software - I love it!

strangesituation_student.zip

When running a cluster analysis, it seems that different versions of JASP yield different results. For example, I was sent a JASP file by a student (attached; data is fictitious) whose results did not match up with the answer sheet in the assignment of our EBCP course at UvA.

Indeed I saw deviating results for running Max. 3 clusters. However, when re-running the analysis I recovered my results (which were in the answer sheet). I run JASP version 0.95.1.

When I re-run the analyses in the version from the UvA software center (v. 0.19.3.0) I get again slightly different results when asking for max 6 clusters (which was the other question in my assignment).

Best wishes, also on behalf of Hilde,
Lucas

Expected Behaviour

The same clustering results across JASP versions.

Steps to Reproduce

1. open the JASP file in v. 0.95.1 and see what results the student had obtained (presumably in an older version of JASP, but this we don't know - I will update you when I find this out).
2. re-run the same analysis
3. check clustering results.

Log (if any)

No response

More Debug Information

No response

Final Checklist

• I have included a screenshot showcasing the issue, if possible.
• I have included a JASP file (zipped) or data file that causes the crash/bug, if applicable.
• I have accurately described the bug, and steps to reproduce it.

[koenderks]

Hi,

What exactly is different? We did fix this issue (#3297) in this pull request (jasp-stats/jaspMachineLearning#423), which impacts the cluster means table and plot. Is your issue related to that? Some elaboration on the exact mismatches is appreciated.

[LucasMolleman]

You can see the differences when following the "Steps to reproduce". It does not seem directly related to the issue you refer to. This is what i see:

before re-analysing in v. 0.95.1 (just loading the JASP file provided by the student)

after re-analysing:

The number of individuals in the cluster changes.

[koenderks]

This might seem weird, but if I refresh your JASP file I get completely the same results as in your first screenshot...
https://github.com/user-attachments/assets/33fee9af-a384-4a62-b767-26ac71254315

[koenderks]

Are you on a different OS than your student?

[LucasMolleman]

for me it DOES update (from screenshot 1 to screenshot 2)! I will try to find out about the OS of the student (who is not responding atm).

[koenderks]

Okay, maybe @vandenman can share his insights into if different OS's can produce different results with the same seed in JASP?

[LucasMolleman]

I just received a reply from the student to my question. Both the student and I work in Windows. We both use JASP version 0.95.1.