Using box plot graphical representations we evaluated the discriminative values of the different summary statistics of flanking markers as a predictor of imputation accuracy for association statistics. Interestingly, the use of the total sum of association values was very robust to efficiently separate true-positive and false-positive association statistics of imputed markers, independent of the size of the sliding window (Figure 5a-c). Using the mean value of the flanking association statistics was only useful in sliding windows of size three. Sliding windows centered in discordant markers (false-positive), as expected, showed consistently lower association statistics than sliding windows centered in true-positive markers. These results were especially impressive when using the total sum of corrected association statistics. The complete set of carried comparison can be inspected in Additional file 7, figure S3 and their application to WTCCC's hypertension results in Additional file 8, figure S4.
