
Dose–exposure and exposure–response models for longitudinal data 55
research has proposed sco re tests for VCs with boundary constraints, but
some require simulations. Practical applications are limited due to difficulties
in implementation in commonly used software.
Now we come back to the test in Section 3.1 where the LR test resulted
in the test statistic χ
2
= 3.33 with 1 d.f. Without considering the boundary
constraint, it is equivalent to a p-value 0.188. However, with the approximate
formula under the constraint, the p-value becomes 0.188/2=0.094.
3.9 Nonparametric and semiparametric models with
random effects
The regression spline approach introduced ...