– The second analysis consists of combining the apple and orange data for a given protocol and condition. This raises the sample size for the comparison to 456 and reduces the effect of differ- ent sensory acuities in each condition, as all the consumers are included (one could argue that if some consumers have different sensory acuity depending on the juice evaluated, the confounding effect would still not be totally eliminated). Results are summarized in Fig. 1.
For all analyses, the resampling d values are larger than their no-resampling counterparts, confirming predictions and previously observed results. The lack of significance for some of the comparisons can most likely be assigned to a lack of power due to an insufficient sample size.