Use gratia::response_derivatives() to get slopes?

[Aariq]

Look into using response_derivatives() from gratia instead of marginaleffects to calculate (average) slopes. I suspect the gratia methods might handle confidence intervals more accurately in terms of testing whether the slopes differ from zero. Need to explore and think on this though.

https://gavinsimpson.github.io/gratia/reference/response_derivatives.html

[Aariq]

I should at least use vcov = vcov(gam, unconditional = TRUE) in marginaleffects::avg_slopes() (#29)

https://stats.stackexchange.com/questions/609685/when-to-use-unconditional-false-in-plot-gam