Utilizing this algorithm as base, Jong et al. [ 10] proposed two methods, ENSEMBLE and JOIN to combine multiple rankings generated by SVM-RFE as in Algorithm 3 and 4. In this paper, we used 25% of training set as hold-out set and used same sets of thresholds and cutoffs as in [ 10], i.e., {0.2, 0.3, 0.4, 0.5, 0.6, 0.7} and {1, 2, 3, 4, 5}.
Even with the same set of data, widely differing threshold estimates can be obtained (Slob 1999).
This is not too surprising, since thresholds were trained/adjusted on the same set of sequences as were searched afterwards employing these thresholds.
Moreover, when the threshold values are high, two PWMs can predict the same set of tentative TFBS; but when score threshold levels are lower, the predicted TFBS sets may be rather different.
Secondly, we identify almost the same set of significant enriched motifs using different thresholds.
In this context, each super-allele might correspond to alternative non-overlapping masking schemes for the same set of variants, for example, different MAF thresholds and/or annotation categories, or to variants in different functional units or genomic regions.
Not all monitoring devices have been evaluated against the same set of criteria, and uncertainty remains regarding acceptance thresholds for the performance of CO monitors and the used reference techniques [ 54- 57].
This marks a marginal improvement compared to the situation reported in 2008, when 54 of the same set of countries had a workforce density below this threshold [ 9].
To test the dependence of the model results on the threshold used to establish if a slope unit contained or was free of landslides, we computed the same set of classification models (LDA, QDA, LR, CM), using different empirical thresholds.
