Optimizing experimental design using the house mouse (Mus musculus L.) as a model for determining grain feeding preferences.

There is little research evaluating flavor preferences among wheat varieties. We previously demonstrated that mice exert very strong preferences when given binary mixtures of wheat varieties. We plan to utilize mice to identify wheat genes associated with flavor, and then relate this back to human preferences. Here we explore the effects of experimental design including the number of days (from 1 to 4) and number of mice (from 2 to 15) in order to identify designs that provide significant statistical inferences while minimizing requirements for labor and animals. When mice expressed a significant preference between 2 wheat varieties, increasing the number of days (for a given number of mice) increased the significance level (decreased P-values) for their preference, as expected, but with diminishing benefit as more days were added. However, increasing the number of mice (for a given number of days) provided a more dramatic log-linear decrease in P-values and thus increased statistical power. In conclusion, when evaluating mouse feeding preferences in binary mixtures of grain, an efficient experimental design would emphasize fewer days rather than fewer animals thus shortening the experiment duration and reducing the overall requirement for labor and animals.