The Standard Error of Measurement (SEM) is an index of measurement error representing the standard deviation of errors attributable to sampling. The SEM provides information about the confidence with which a particular score can be interpreted, relative to a single individual's true score. Thus, a small SEM represents greater confidence that a score reflects the individual's true performance and skill level. The SEM is based on the formula: SEM = Sd x the sq root of r, where Sd = observed Standard Deviation, and r = Reliability.
The SEM, in most cases, is student and measure-specific. All IRT-based measures begin on the logit scale, which has an M=0, SD=1. Each vendor then creates its unique scale. The SEM of our scales must be below .20 to satisfy our standards for reliability.