Hi, I'm trying to run a GAM regression in mgcv with: 1. My own knots(which I learned to do here https://stat.ethz.ch/pipermail/r-help/2012-December/342723.html) 2. One of the cubic splines is defined as monotonic increasing This post(r.789695.n4.nabble.com/Monotone-splines-td833854.html) suggests using mono.con, but that's run after the call to the gam function so it doesn't return me a gam object( http://127.0.0.1:11821/library/mgcv/html/gamObject.html) with all the detailed summary that I need I see there is a package scam( http://cran.r-project.org/web/packages/scam/index.html) that returns a gam object and conveniently lets me define a spline with monotonicity constraints(bs = "mpi" or bs = "mpd") BUT the scam function doesn't have an optional knots parameter like gam does. Any suggestions how I can run a regression with both of these things and get a gam object? Thanks, Nadav [[alternative HTML version deleted]]