
NONPARAMETRIC POISSON MODEL 107
• semiparametric additive regression,
ln
λ
= s
0
+ s
1
x
1
+ ···+ s
i
x
i
+ s
i+1
(x
i+1
) + ···+ s
k
(x
k
)
where s
0
,...,s
i
are constants, and s
i+1
(·),. . . , s
k
(·) are smoothers.
Splines were first applied to model the rate in a Poisson regression in 1986 by an Irish
statistician Finbarr O’Sullivan (1958-), and American statisticians Brian S. Yandell
(1952-), and William J. Raynor Jr. (1949-).
1
5.3.2 SAS Implementation
The syntax for fitting a nonparametric (or semiparametric) Poisson model is similar
to the one for the nonparametric (or semiparametric) logistic model with adjusted
specifications for the link function and the probability distribution. ...