InfraRed Imaging Spectrograph (IRIS) is one of the three instruments that will be available at the first light of TMT. Working with the facility adaptive optics system NFIRAOS, IRIS provides diffraction limited imaging and integral field spectroscopic capabilities in the near-infrared wavelength range. One of the unique observing performances of IRIS is precision relative astrometry at a level of 30 uas. Achieving this unprecedented accuracy necessitates a number of contributing error terms, including target property errors, atmospheric property errors, AO performance errors, optomechanical errors, detector characteristic errors, and data reduction errors, to be reduced by design, calibration, operation, and data reduction. We have created a detailed astrometry error budget which identifies all error terms, quantifies them, and distributes allocations to individual terms to enable the required astrometry accuracy. This error budget is written in terms of astrometric accuracy, i.e., uas, and each error term needs to be translated to technical languages through analyses and simulations when it flows down to system and subsystem requirements. This paper describes the analyses and simulations that we have done to derive the requirements for IRIS with a focus on atmospheric dispersion, optical distortion, and detector characteristics.