
248 Exposure-Response Modeling: Methods and Practical Implementation
in this chapter, should be implemented. If the y-part model is no nlinear or
fixed u
i
cannot be fitted, it is prudent to apply an adjustment at least as a
sensitivity analysis.
TABLE 8. 1
Summary of
ˆ
βs and their SEs es tima ted us ing either the linear mixed model
or a linear model with fixed u
i
.
Time pattern
ˆ
β(SE) (Mixed model)
ˆ
β(SE) (Fixed u
i
)
(1,2,4,8) 0.98 (0.0024) 1.00(0.0025)
(1,2,4) 0.89 (0.0044) 1.00(0.0056)
(1,4) 0.85 (0.0052) 1.00(0.0068)
(1,2) 0.78 (0.0052) 0.97(0.0257)
Even with some study desig ns the naive estimates themselves may not be
robust enough; these designs may create ...