jasp-stats / jasp-issues

This repository is solely meant for reporting of bugs, feature requests and other issues in JASP.
59 stars 29 forks source link

[Bug]: Bootstrapping on Classic Network Anaysis Fails on Windows #2863

Open sitnok67155 opened 3 months ago

sitnok67155 commented 3 months ago

JASP Version

0.19.0

Commit ID

No response

JASP Module

Network

What analysis are you seeing the problem on?

Classic Network Analysis

What OS are you seeing the problem on?

Windows 11

Bug Description

When attempting to use boostrapping (at any type and at any number of bootstraps), I receive this error:

image

Another person ran into this problem and reported it on May 6th but I do not believe it has been resolved.

Expected Behaviour

The network should have been bootstrapped

Steps to Reproduce

1.classical network analysis

  1. added the variables
  2. ticked all plots and tables
  3. ticked boostrap network, tried 10/100/1000 number of bootstraps, and tried nonparametric and case bootstrap types. ...

Real Final Data.csv

Log (if any)

No response

More Debug Information

No response

Final Checklist

tomtomme commented 2 months ago

confirmed with 0.19 flatpak and also 10000 bootstraps and also with a subset of the variables. was also present in 0.18.3 according to #2704

Parametric bootstraps etc. works. Only nonparametric and case type are affected with this data set. Data from our library works with every option.

jasp file: Real.Final.Data.jasp.zip

tomtomme commented 1 month ago

We now have multiple reports since JASP 0.16.4 with this error over multiple datasets on win10, win11, MacOS Silicon and flatpak, all confirmed. So I change this to critical.

2917

2863 (this one)

2704 (2 independent reports)

2453

2088

2063

1902

The last contained a fix that should display a better error message, but it might have added it only for the correlation matrix error not for this non-parametric bootstrap related error... which seem very similar to me at least :D

@vandenman, you wrote in #2088:

Thanks for reporting the issue! What goes wrong internally is that the correlation matrix for the second group (Who 5 days == 2) is not positive definite. If you change the correlation method to Cor or the method for Missing values to Exclude listwise then the error disappears and you do get output. I'll adjust the code so you receive a more informative error message in the future.

And this was the pull that tried to fix it: https://github.com/jasp-stats/jaspNetwork/pull/86

tomtomme commented 2 weeks ago

Still there with 0.19.2 beta flatpak.