Removing Bias — The SIMEX Procedure

SIMEX (Simulation-Extrapolation) is a very general technique that helps to correct for bias in estimates caused by errors in measurements of predictors. The method is well established in statistical practice, but seems to not be as widely known in actuarial circles. Using ordinary least squares regression as an example, the method is illustrated using some simple R code. The reader is encouraged to run the code (which is highlighted in grey) in R themselves to get hands-on experience using the method.

Keywords: SIMEX, MCSIMEX, Naïve Estimate, Unbiasedness, Bias