I am trying to use SAS PROC ADAPTIVEREG to do a variable selection and search for optimal knot position for spline transformation. One question that I encounter is how to, if possible, restrict the maximum number of knots for each individual variable. For example, I am doing a logistic regression with the response variable y and 10 predictors (x1 to x2). In the search, I would like to put a restriction that at most 2 interior knots can be used for x1. I know how to set the maximum number of basis functions for all variables via MAXBASis= option in MODEL statement. But not clear on how to do it for the individual variables.
Question: Is there a way to do it? If so, how to do it.
