Animal Study Workflow

In vivo animal studies must now comply with ARRIVE 2.0 reporting guidelines — mandatory for Nature, Science, PLOS Biology, and most high-impact journals. Licklider generates an ARRIVE 2.0-aligned Methods text, enforces the pseudoreplication guard for nested tissue samples, logs all exclusions, and produces a power justification for the ethics submission.

STEP 1 — The Scenario

Researcher Profile

A pharmacologist testing the effect of a novel anti-obesity compound on liver steatosis in a diet-induced obesity (DIO) mouse model. Experiment design: 3 groups (Vehicle, Low Dose, High Dose), 10 mice per group. At sacrifice, 3 liver slices per mouse are analyzed for lipid content (% area staining) by Oil Red O histology.

Research Question

"Does the compound reduce liver lipid content in DIO mice, and is the effect dose-dependent?"

Dataset

30 mice: Vehicle (n=10), Low Dose (n=10), High Dose (n=10)
3 liver slices per mouse = 90 total histological measurements
Outcome: Liver lipid content (% Oil Red O staining area), compositional
Columns: MouseID, Group, Slice (1/2/3), LipidPct

STEP 2 — Licklider Analysis Path

1. Pre-Study Power Calculation (Ethics Stage)

Before the experiment, the researcher uses Licklider to calculate the required sample size. Based on pilot data (5 mice/group): expected mean difference between Vehicle and High Dose = 15%, SD = 8%. Cohen's d ≈ 1.875. Required n per group for 80% power, α=0.05: n=5. With 20% attrition adjustment: n=7 per group. The researcher uses n=10 to account for uncertainty in the pilot estimate and to increase power to ~95%. Power justification statement generated.

2. Data Upload & Replicate Structure Detection

Licklider detects the MouseID column (30 unique values) and the Slice column (3 levels). The pseudoreplication guard fires: "3 slices per mouse are technical replicates — analysis must be at the mouse level (n=10 per group), not the slice level (n=30 per group)."

3. Technical Replicate Averaging

The 3 slice measurements per mouse are averaged → 30 mouse-level observations (10 per group). Analysis proceeds on mouse-level data. Averaging logged in the Preprocessing Audit Log.

4. Exclusion Handling

During the experiment, 2 mice (1 Vehicle, 1 Low Dose) died before sacrifice — not treatment-related. The researcher records these in Licklider's exclusion log with the reason (spontaneous death, animal welfare). Final n: Vehicle=9, Low Dose=9, High Dose=10. The N Disclosure (1.2.1) guard updates all disclosures: input N=30, analysis N=28, 2 excluded (reason: death, pre-specified exclusion criterion).

5. Statistical Analysis

Compositional data check: LipidPct is bounded [0, 100]. Arcsine-square-root transformation applied. One-way ANOVA on transformed data (3 groups, unequal n) with Holm-corrected pairwise post-hoc. Effect size: η².

6. Figure Generation

Box plot with individual mouse-level data points (not slice-level). Groups on x-axis, lipid % on y-axis (original scale). Significance brackets with Holm-corrected p-values. Representative histology images as a separate panel (mock placeholder shown).

STEP 3 — Guards & Disclosures Activated

Pseudoreplication Detection (4.5.1)

The slice-level data (n=90) cannot be used as the unit of analysis. Licklider fires the guard and enforces mouse-level averaging. The n reported in all outputs is n=28 mice (not 84 slices).

Outlier Exclusion Log (1.1.2)

The 2 excluded mice are logged with: MouseID, Group, Reason (death), Date of exclusion, Whether exclusion criterion was pre-specified. The log is included in the export package.

N Disclosure (1.2.1)

Input N=30 → Analysis N=28 → Excluded=2 (2 deaths). Per-group n: Vehicle=9, Low Dose=9, High Dose=10. All disclosed in figure legend and Methods text.

Compositional Data Warning (4.7.2)

LipidPct is proportion data. Arcsine-square-root transformation is applied and disclosed. The figure axis label includes "% area (arcsin-transformed for analysis)".

Sample Size Justification (4.7.3)

The power justification from the pre-study calculation (n=10 per group, expected power ~95% based on conservative effect size from pilot) is locked into the export and appears in the Methods text.

ARRIVE 2.0 Compliance Checklist

Licklider generates a pre-filled ARRIVE 2.0 checklist from the Data Contract, exclusion log, power calculation, and analysis decisions. Unfilled items are flagged for manual completion.

STEP 4 — Export Package

Figure: Mouse-level box + strip plot — Vehicle, Low Dose, High Dose — 300 dpi TIFF/PDF
Statistical Report: Averaging summary (3 slices → 1 mouse mean); ANOVA table with transformation disclosure; Holm-corrected post-hoc table; η² (95% CI); per-group n; exclusion summary
Methods Text (ARRIVE 2.0-aligned): "Liver lipid content (% Oil Red O area) was quantified from n=3 histological sections per animal, averaged to a single mouse-level value prior to analysis. Two animals were excluded due to spontaneous death (pre-specified exclusion criterion). Final n: Vehicle=9, Low Dose=9, High Dose=10. Data were arcsine-square-root transformed before one-way ANOVA with Holm-corrected pairwise post-hoc. Sample size was determined by a priori power analysis based on pilot data (Cohen's d=1.875, α=0.05, target power ≥80%). Effect size is reported as η² (95% CI)."
Exclusion Log: Per-animal exclusion record with reason and pre-specification status
Power Justification Statement: Full power calculation with inputs, output n, and attrition adjustment
ARRIVE 2.0 Checklist: Pre-filled where Licklider data is available; flagged items requiring manual completion
Preprocessing Audit Log: Slice averaging; transformation; exclusion events