I suspect you'll get more responses if you show your PHREG code, the R code, and provide sample data. Also, let us know what the predictions of the R package are for a few example values.    Since you claim the result is "close-ish," the issue could be the way you are specifying the knots. Please make sure that you are using B-splines and the same number of knots that the R packages uses. You might need to use the knotsWCE() function in the WCE package to make sure that SAS and R are putting knots at the same locations. Also, B-splines require external knots. The positions of the external knots determine the basis functions, so you need to know where R is placing the external knots. The SAS documentation includes a section that includes graphs that show a B-spline basis for various choices of knot placement: https://go.documentation.sas.com/doc/en/statcdc/14.2/statug/statug_introcom_sect023.htm  ... View more

I love the idea of basing a paper on the PROC GLM paper from 1976!   If we're wandering down memory lane, here's a picture of me with my SAS 1979 manual (the last time the SAS documentation fit into one volume)     and an example of the "high res" graphics in it from PROC GLM. Who says you have to know ODS?     Maybe we need an "old codgers" forum, where us old-timers can reminisce about punched cards and IBM Selectrics!   Best,    Tom       ... View more

A tiny fraction of what you can do today with GLM existed in 1976. That is far more relevant than a 1976 program working today. GLM was rewritten and enhanced by several developers, and much of the functionality has been subsumed by newer procedures such as GLMSELECT. ... View more