
238 Exposure-Response Modeling: Methods and Practical Implementation
fo cus is on nonlinear models because for these models simulation approa ches
are often time consuming, and may lead to failure of fitting the model in a con-
siderable number of runs in the simulation. The approach introduced here is
easy to use, although it only gives asymptotic bias estimates. This is often suf-
ficient for a sensitivity analysis. Closely related to this topic is the assessment
of ignored covariates in regression analy sis, but its fo cus is on nonconfounding
covariates.
For illustration, we start from the linear models (8.1), where the confound-
ing bias in the