32 Exposure-Response Modeling: Methods and Practical Implementation
other approaches that use more information in the data in later par ts of the
book.
Using spline functions in SAS is also easy for some proce dures such as proc
GLIMMIX. The EFFECT statement
EFFECT bs = spline(exposure / knotmethod=percentiles(3));
creates B-spline function effects with two knots at the lower and upper bo unds
and internal knots at the median of the e xposure data. Then in the model
statement the effects bs are included as covariates. This statement can be
used in a number of other SAS procedures.
As a general approach, creating B-spline functions as covariates outside of
a procedure or function for model fitting can be used in almost all s oftware
including R/Splus functions ...