Table 6. The robustness of each pooling method to over- and underestimation of the number of positive samples.

	No. of Samples	DNA Sudoku	2D Pooling	S-Stage	Halving	Gen. Binary Splitting	Modified 3-Stage
Change in No. Tests (and Steps) with Overestimate of Positive Samples (k = 1, $\widehat{k}=20$)	96	+935	+45	+22 (-3 steps)	0	+68.77 (+32.19 steps)	+21 (-1 step)
	384	+1187	+217	+37.41 (-3 steps)	0	+109.30 (+73.08 steps)	+31.89
	1,536	+1455	+945	+35.39 (-3 steps)	0	+146.96 (+110.89 steps)	+54.65
Change in No. Tests (and Steps) with Underestimate of Positive Samples (k = 20, $\widehat{k}=1$)	96	-857.68 (+1 step)	+11.5 (+1 step)	+14.53 (+3 steps)	0	+61.99 (+62.57 steps)	+11.75 (+1 step)
	384	-1027.82 (+1 step)	-60.79 (+1 step)	+11.34 (+3 steps)	0	+58.40 (+58.40 steps)	+54.74
	1,536	-1212.81 (+1 step)	-691.86 (+1 step)	+7.34 (+3 steps)	0	+56.80 (+56.80 steps)	+108.85

Results show the average increase/decrease in the number of tests and steps when k = 1 and $\widehat{k}=20$ (overestimation) and when k = 20 and $\widehat{k}=1$ (underestimation). The most and least robust methods in each row are indicated in green and orange, respectively.