Default behavior should be that cnr looks for the smallest influence weighted knot from all marginals and removes that knot.
However, the user should be able to specify which marginals are to be considered for the reduction. Say a 3D tensor surface, built on x1, x2, and x3, is given and the user knows/wants to retain the knots specified for x2. cnr needs an option so that the algorithm will only run on x1 and x3.
Possible API: cnr(a_cpr_cn_object, margin = c(1, 3))
Default behavior should be that
cnr
looks for the smallest influence weighted knot from all marginals and removes that knot.However, the user should be able to specify which marginals are to be considered for the reduction. Say a 3D tensor surface, built on
x1
,x2
, andx3
, is given and the user knows/wants to retain the knots specified forx2
.cnr
needs an option so that the algorithm will only run onx1
andx3
.Possible API:
cnr(a_cpr_cn_object, margin = c(1, 3))