Calculator

Enter your historical control data and study parameters to get a concrete recommendation.

A · Your Historical Control Data

Enter summary statistics from one specific type of control group across your previous studies. Each row is one study — enter only the animals from the control group that matches exactly what you plan to use in the new study.

Only matching controls. If your new study uses a vehicle control (e.g. NaCl i.p.), enter only vehicle control data from previous studies — not untreated controls, not sham controls, not positive controls.
Presets:
# n (animals) Mean SD
B · Heterogeneity Assessment

How consistent are your historical controls? The DerSimonian-Laird method estimates the between-study standard deviation (τ).

—
Between-study SD (τ)
—
I² heterogeneity
—
Q statistic (df=?)
Sensitivity: ESS across heterogeneity levels
High borrowing Moderate Negligible Arrow = your τ
C · Effective Sample Size

How many of your historical animals effectively contribute to the new study? Think of it as: "my N historical animals are worth ? concurrent animals."

—
Historical animals
—
neff (raw)
—
neff (robust, w=20%)
—
Information loss
w = 0.20
0 = trust historical data fully. 0.2 = recommended (Schmidli et al. 2014). 0.5 = maximum caution.
D · New Study Design

Configure your planned study. Choose the design type and specify the effect size.

Advanced Settings
E · Recommendation

Design Comparison

DesignControlsTreatmentTotalAnimals saved
F · Results
G · Expert Panel

Explore how the recommendation changes as parameters vary.

Sensitivity Analysis

2.00
0.20
5
—
ESSrobust
—
nconcurrent
—
Animals saved
H · What Happens After Planning?

This calculator handles Step 1: planning — determining how many concurrent control animals your study needs. When the study is complete, the historical borrowing must also be part of the statistical analysis.

Important distinction. If you reduce your control group based on this tool but then analyse the data with a standard t-test, you will be underpowered. The reduced control group was designed assuming historical data would be formally incorporated into the analysis.
Step 1 — Planning (this tool)
  • Estimate τ from historical data
  • Compute ESS (the discount)
  • Determine reduced control group size
  • Document in ethics application
No special software needed
Step 2 — Analysis (after the study)
  • Fit the MAP prior via Bayesian software
  • Update with your new concurrent data
  • Run the formal Bayesian test
  • Report posterior & credible intervals
Requires a statistician

See Methods for the full statistical framework.