Splines with difference penalty and monotonic constraints

[lorentzenchr]

As a user of GLMs/GAMs, I would like to

• splines with a difference penalty without the hassle to specify the P2 matrix myself
• optimize a single penalty per term (think of 2 features both modelled as splines)
• use splines with monotonicity constraints

About monotonic splines, cf. Ramsay "Monotone Spline Regressions in Action" https://doi.org/10.1214/ss/1177012761:

• One can use I-spline basis; or
• Use differences of B-Splines
• One needs to restrict the coefficients to be either >=0 (increasing) or <=0 (decreasing)