In this section we describe a simulation study that was designed to assess the bias and mean- squared error of ˆ under different basis representations for f and the five approaches to selecting df that were described in Section 2. Our goal is to generate data from confounding scenarios that are comparable with situations found in real data and to evaluate the estimation procedures in each of these scenarios. The definition of the scenarios relies on the timescales at which con- founding occurs and the strength of the concurvity between the pollutant series and the seasonal trend. All the simulations were conducted in R by using.
