This document discusses output analysis for simulation models and eliminating initial bias. It explains that initializing a simulation with an empty or idle state can cause initialization bias, skewing early results. Two common approaches to address this are to start simulations in a more representative state or remove an initial warmup phase of results. The ideal is to start from the true steady state distribution, while typically the warmup phase is eliminated by restarting after a pilot run determines how long the initial bias persists.