Would You Agree If N Is Three? On Statistical Inference for Small N

Area of Science:

• Animal research methodology
• Statistical inference in biology

Background:

• Traditional nonhuman primate studies often use small sample sizes (2-3 animals).
• Previous work suggested using 1 animal for sample inference or 5+ for population inference.
• A recent framework proposed using 3 animals and majority outcome for population representation.

Purpose of the Study:

• To evaluate the error rate of a 3-animal testing framework across varying outcome typicalities.
• To determine the conditions under which small sample sizes (2-3 animals) provide useful population inference.
• To assess the inferential value of majority outcomes from small nonhuman primate study groups.

Main Methods:

• Statistical analysis of a proposed 3-animal testing framework.
• Simulation under diverse typicality assumptions for the representative outcome.
• Conjunction analysis to evaluate inferential bounds from 2-3 animal outcomes.

Main Results:

• The framework's error rate is highly sensitive to outcome typicality.
• Acceptable error rates necessitate very high typicality (>87%), questioning the need for >1 animal.
• Increasing sample size from 1 to 3 animals primarily benefits typicality values of 70%-90%.

Conclusions:

• The 3-animal framework's validity is limited by outcome typicality.
• Small sample sizes (2-3 animals) provide limited population inference, with a low inferred typicality lower bound (9%).
• Reporting the inferred lower bound of typicality is recommended if 2-3 animals are used.