Consider using less stringent convergence tests

[singmann]

Following Doug Bates recommendations, it might be best to disable gradient-based tests in fitted mixed models by using control = lmerControl(calc.derivs = FALSE). It speeds up testing and these tests are rather likely to produce false positives.

[mattansb]

@singmann do you have a reference on the false-positives point? (I've seen the speedup recommendation in the lme4 vignette)

[singmann]

Unfortunately nothing citable. Doug Bates said this in person during SMLP 2024.