For the smallest d0, a protocol feature could have played a role in the somewhat counterintuitive trend observed: the size of the estimated d0 value variance. For a sample size of 228 and a d0 of 0.44, the tetrad d0 variance is 0.043. For the same sample size and the same d0 value of 0.44, the triangle d0 variance would be 0.16. Therefore, even if the triangle testing would generally result in a slightly higher d0 value due to more limited experimental variance (as seen in the other three conditions), the imprecision of the d0 estimate with its larger variance negates that potential advantage. Another example of this effect is visible in Table 1, where it can be seen that in order for the triangle d0 to have approximately the same variance as that of the 0.44 tetrad d0 (variance = 0.043), it needs to be more than twice as large (apple juice resampling condition, d0 = 0.90, variance = 0.05). This illustrates a further benefit of the tetrad over the triangle, namely the higher precision in the estimation of the size of the underlying difference between the products.
Nevertheless, the expected