To further explore the relationship between association statistics derived from imputed and from directly genotyped allelic frequencies, we sorted and determined the top ten markers that would be considered strongly associated based on the evidence provided solely by imputed frequencies (Table 3). This ranking analysis is extremely useful for the determination of markers strongly associated with a phenotype and avoids the error caused by an inadequate selection of a significance threshold. The analysis of the results organized in Table 3 highlight that association statistics for top-associated SNPs derived from imputed frequencies are highly inflated in comparison to their empiric counterparts. A huge proportion of these imputed markers (9 out of 10) would be considered not associated to the phenotype if evidence provided by direct genotyped allelic frequencies were to be used. It's an interesting fact that all these biased markers share similar frequencies in their minor alleles, all very close to 0,5, this allelic condition was further explored in this manuscript.
