100 Exposure-Response Modeling: Methods and Practical Implementation
and note that the assumption leads to a simple solution to the problem, but
may not be reasonable for some practical scenarios. These models are exactly
the same as models (4.66) if g(c
∗
i
, β) is linear, although the notation is dif-
ferent, with c
i
and d
i
replaced with c
∗
i
and c
i
, respectively. Therefore, the
same RC approaches for models (4.66), such as the weighted LS appr oaches
described above, apply here. For example, for a linear y-part model, the RC
approach first fits model c
∗
i
= h(c
i
, θ) + e
i
and estimate E(c
∗
|c
i
) by h(c
i
,
ˆ
θ).
Then a weighted LS can be used to estimate β.
4.7.3 Using a surrogate for C
max
: An example
Here we c onsider a practical example of using a surroga te for ...