Every example below uses the same study design parameters: a two-group comparison with effect size δ = 10 g, within-study SD σ = 15 g, α = 0.05, power = 80%, robustification weight w = 0.2, and a minimum of 5 concurrent controls (never zero). The endpoint is body weight in grams. All values are computed live from the study data using the same statistical engine as the Calculator.
Dr. Ratzinger runs a toxicology lab in Hannover. Over the past three years her team has completed eight regulatory studies, all following the same protocol to the letter: male Wistar rats, 8 weeks old at study start, sourced from the same breeder colony. The same senior technician, Frau Pipetti, has handled every animal. The vehicle control is always 0.9% NaCl administered by intraperitoneal injection. The endpoint is terminal body weight.
The result of this discipline? Eight control group means that cluster tightly around 250 grams. This is what consistency looks like in practice.
Study Timeline — Control Group Means
The Borrowing Funnel — How Many Animals Actually Count?
Impact on Your Next Study
Now imagine Dr. Ratzinger's lab four years later. Frau Pipetti retired after the third study. Her replacement, a capable but less experienced technician, handles animals slightly differently — perhaps holding them a bit longer during weighing, or conducting measurements at a different hour. One batch of animals also came from a different breeder cohort when the usual supplier had a shortage.
The protocol on paper is identical. The vehicle control is still NaCl i.p., the rats are still Wistar males. But the means have started to wander: 240 g in one study, 258 g in another. Is this enough drift to undermine historical borrowing?
Study Timeline — Control Group Means
Impact on Your Next Study
A European research consortium wants to reduce animal use across its member laboratories. Someone suggests pooling vehicle control data from three sites — Hannover, Lyon, and Uppsala — to build a large historical control database. On paper, all three use "Wistar rats" and "NaCl vehicle controls." The idea sounds efficient.
But "Wistar" from breeder A in Germany is not the same sub-strain as "Wistar" from breeder B in France. Housing temperature differs by 2°C. The Swedish site uses a different feed supplier entirely. Animal ages at study start range from 8 to 12 weeks. Control group means span from 228 g to 270 g — a 42-gram spread. This is not one population. It is three.
Study Timeline — Control Group Means (Multi-Site)
Impact on Your Next Study
What If You Ignored the Heterogeneity?