To determine if the bias between association statistics could be predicted by common filtering criteria, we used a graphical representation plotting the observed bias between association statistics against calling probabilities, MAF and Hardy-Weinberg equilibrium of each marker in the dataset (Figure 3). The results of the plotted figure suggest that the Hardy Weinberg deviation, as expected, cannot be used as a predictive variable since the most prominent bias were encountered in markers that showed only a moderate deviation from equilibrium. The same procedure was applied to the empiric and imputed calling criteria, in both analysis the analyzed features do not show any predictive value, since the highly biased association values were concentrated in high quality empiric markers and randomly distributed for imputed markers. These features were useful for filtering highly biased markers in the creation of the filtered dataset but different thresholds of calling probabilities were not efficient predictors any further. When the same plotting routine was used using MAF as a predictor for bias (Figure 3), interesting results could be observed. Most prominent bias was encountered in markers with
