Pre-Study Power Calculation

Sample size calculation requires three inputs: effect size, alpha, and desired power. Licklider asks the researcher to justify each input, calculates the required n, and produces a Methods-ready statement that satisfies ethics committee and grant reviewer requirements — including a sensitivity table showing n across a range of assumptions.

STEP 1 — The Scenario

Researcher Profile

A PhD student planning an in vivo mouse experiment to test whether a novel peptide reduces fasting blood glucose in a diet-induced obesity model. Pilot data from 5 animals per group shows a mean difference of 25 mg/dL with a pooled SD of 20 mg/dL.

Research Question

"How many mice per group do I need to detect a 25 mg/dL difference in fasting blood glucose with 80% power at a two-sided α of 0.05?"

Inputs Available

• Expected mean difference: 25 mg/dL (from pilot data)
• Pooled SD: 20 mg/dL (from pilot data)
• Alpha: 0.05 (two-sided, standard)
• Desired power: 0.80 (80%)
• Planned test: Two-sample Welch's t-test (independent groups)

STEP 2 — Licklider Analysis Path

1. Effect Size Calculation

Licklider calculates Cohen's d from the provided inputs: d = |µ₁ − µ₂| / SD = 25/20 = 1.25. This represents a large effect. Licklider notes: "A Cohen's d of 1.25 is large. If the pilot estimate is uncertain (n=5), consider a more conservative effect size for power calculation. A sensitivity analysis across d=0.8, 1.0, 1.25 is recommended."

2. Sample Size Calculation

Using the two-sample t-test formula (G*Power equivalent), Licklider calculates the required n per group for:

• d=1.25, α=0.05, power=0.80 → n=12 per group
• d=1.00, α=0.05, power=0.80 → n=17 per group
• d=0.80, α=0.05, power=0.80 → n=26 per group

3. Sensitivity Table

A sensitivity table is generated showing required n across a range of effect sizes (d=0.5 to 1.5), power levels (70%, 80%, 90%), and alpha levels (0.01, 0.05). The researcher selects the scenario that reflects the most defensible assumptions for their ethics submission.

4. Attrition Adjustment

If the researcher expects 10% animal attrition (deaths, exclusions), Licklider inflates the required n by 1/(1-0.10) = +10%. Final recommendation: n=13–14 per group (rounded up from 12 × 1.1).

5. Methods Statement Generation

Licklider generates a ready-to-paste Methods section entry that discloses all assumptions: expected mean difference, SD, Cohen's d, α, power, test type, and the attrition adjustment.

STEP 3 — Guards & Disclosures Activated

Power & Sample Size Disclosure (4.1.4)

Licklider requires disclosure of all four power calculation inputs: effect size (and its source), alpha, power, and planned test. A sample size stated without these supporting details cannot be exported in a Methods-ready format.

Effect Size Source Enforcement

Licklider asks: "What is the source of your effect size estimate?" and records the answer. Options: pilot data (with n), published literature (with citation), clinical meaningful difference (with justification), or assumed. "Assumed" triggers a warning: "Power calculations based on assumed effect sizes are common but may underpower studies if the true effect is smaller."

Sample Size Justification (4.7.3)

The Sample Size Justification guard confirms that the final recommended n is consistent with the disclosed inputs. If the researcher modifies the n downward (e.g., due to animal welfare constraints), the resulting achieved power is calculated and disclosed.

STEP 4 — Export Package

• Sensitivity Table Figure: n per group across effect size × power scenarios — 300 dpi TIFF/PDF
• Power Calculation Report: All inputs (mean difference, SD, Cohen's d, α, power, test); required n per group; attrition-adjusted n; achieved power at selected n
• Methods Text (auto-generated): "Sample size was calculated using G*Power-equivalent two-sample t-test power analysis. Based on pilot data (n=5 per group), we estimated a mean difference of 25 mg/dL (SD=20 mg/dL; Cohen's d=1.25). Assuming α=0.05 (two-sided) and 80% power, a minimum of 12 animals per group is required. Accounting for 10% expected attrition, we will enroll 14 animals per group."
• Sensitivity Table: Full n-per-group matrix for reference