In a complementary way, we used a graphical representation to highlight this feature plotting the corrected association of statistics of sliding windows centered in true-positive and false-positive imputed markers considered associated in human chromosome 10 (Figure 6). In the upper left panel, a small region was highlighted showing three totally isolated imputed markers that could be considered associated to the phenotype of interest, but not by their empiric-derived statistic disfavoring the true/false positive ratio. When the same procedure was carried in imputed markers that were concordant to their empiric measures (upper right panel), it is noteworthy that a considerable proportion of markers weakly associated to the phenotype is also clustered in the same small chromosomal region. This result suggests that a preliminary analysis of surrounding markers could be used to flag and identify imputed markers that do not reflect the true empiric frequencies and could erroneously be considered associated.
