An important consideration in the estimation of LD Score is the extent to which the sample from which we estimate LD Score matches the sample for the association study. If there is mismatch between LD Scores from the reference population and the target population used for GWAS, then LD Score regression can be biased in two ways. First, if LD Scores in the reference population are equal to LD Scores in the target population plus mean-zero noise, then the intercept will be biased upwards and the slope downwards. This is conceptually equivalent to increasing the measurement error of LD Score. Secondly and perhaps more importantly, consider the scenario where there is a directional bias in average LD Score such that the LD Scores in the reference population are systematically higher or lower than in the target population. Under such a scenario, then the LD Score regression intercept will be biased downwards or upwards, respectively (Online Methods).
