I‘ll answer here github.com/florianharti... , feel free to add to the discussion

That being said, I agree that standard S3 functions schools conform to R defaults as far as possible and I’ll have a look at the example you point out … there are a number of other standard S3 functions that we still need to implement

Hi Norm, thanks for your comments - if you encounter any bugs or have suggestions for changes, please post them at github.com/citoverse/ci...

deep neural networks in one line of R code, with convenient options for automatic hyperparameter tuning, xAI and bootstrapping.

The current CRAN version supports only fully connected DNNs, but our development version on GitHub also includes CNNs and multi-modal architectures.

Nice! How do you see this in relation to ggeffects?

All details and code in arxiv.org/abs/2411.15105

But why does the effect appear in the first place? We believe the reason is that as mutualism rates are predicted by latitude, mutualism and latitude are perfectly nonlinear confounded. Using only a linear model, nonlinear effects of latitude create spurious effects of mutualism.

The main issue we see is that the mutualism covariate is not measured, but predicted by a very simple model that only depends on latitude (R2 = 0.14). Predicting mutualism rates with a more precise model (R=0.39) removes any effect of this covariate in the main regression.

Well, I've heard it said that great minds think alike ... ;)