same reference haplotypes from the HapMap projects, there could be an intrinsic bias toward positive replications and initial results based on imputed data should be followed by actual genotyping of the missing data to confirm real effects. The efficacy of imputation of data from populations that are not well represented in the HapMap project, such as African Americans, is still an open question [58], and similar caution should be used when imputed data of these populations are used for meta-analysis.
