[jaspMixedModels] Fix parametric bootstrap error in Generalized Linear Mixed Models. ...

[sisyphus-jasp]

Summary

PR Summary: Fix parametric bootstrap error in Generalized Linear Mixed Models

Root Cause

The afex::mixed function with method="PB" (parametric bootstrap) internally calls lme4::glmer with REML=FALSE. However, glmer() does not accept the REML argument because GLMMs are always fit using ML (maximum likelihood), not REML. This caused the error: "unused argument (REML = FALSE)" when users tried to use parametric bootstrap with GLMMs.

Fix

Added an .onLoad hook in R/MixedModelsCommon.R that patches lme4::glmer to accept but ignore the REML argument. The patch:

1. Uses unlockBinding to modify the lme4 namespace
2. Creates a wrapper function that removes the REML argument before calling the original glmer
3. Uses lockBinding to restore the lock
4. Only patches once using a flag to avoid issues with repeated loading during tests

Test Results

• All tests pass: [ FAIL 0 | WARN 0 | SKIP 3 | PASS 52 ]
• This matches the baseline results, confirming no regressions

Caveats

• The patch modifies lme4's namespace at runtime
• The skipped tests are unrelated to this fix (they're skipped due to afex issue Translations update from Hosted Weblate #134 and ggplot2 incompatibility)

Implementation Plan

No plan provided by agent.

Test Results

Baseline (before fix): [ FAIL 0 | WARN 0 | SKIP 3 | PASS 52 ]
After fix: [ FAIL 0 | WARN 0 | SKIP 3 | PASS 52 ]

Metadata

• Upstream baseline: e95dd7c -- CI: failing
• Model used: minimax/minimax-m2.5
• Job ID: 1771331714
• Generated by: OpenClaw autonomous coding agent

---

🤖 This PR was automatically generated by the OpenClaw agent. Please review carefully before merging.