Use when defining n, choosing statistical tests, correcting for multiple comparisons, and reporting error bars for a Cancer Cell (Cell Press) manuscript. Focuses on biological statistics and avoiding pseudo-replication; it does not design experiments or build figures.
SkillScore88Maintained
Automatic
◆Research pack
Last verified 8/31/2026
What it changes
Passed 21 of 24 test cases with the skill, and 18 of 24 without it.
Without skill75%
With skill88%
Checks it passed that the model alone missed
Recommends cc-figures-tables in the next step block
Includes the header 【n definition】 in the assessment
Includes the header 【Multiplicity】 in the assessment
and 4 more in the full report
5 fixed16 passed either way1 still failing2 broke
Pass-rate lift
+13 pp
Tokens
+34%
Agent turns
0%
Response time
-19%
Model gemini-3.6-flashMeasured 8/31/2026v121 of 24 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 5
Prompt
Before finalizing submission advice on volatile journal policies, what specific reference map document must be re-opened to confirm compliance floor details?
What was checked, without and with the skill
Checked
Without skill
With skill
Names resources/official-source-map.md as the file to re-open
Missed
Passed
case-17 · gemini-3.6-flash · 8/31/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md89 linessingle file
Biological Statistics (cc-statistics)
When to trigger
n is ambiguous, or you suspect pseudo-replication
Unsure which test fits the data and design
Many comparisons without multiplicity correction
Error bars / variability are unlabeled in figures or legends
Defining n (the core issue)
n = number of independent biological replicates (separate mice, independent cultures/passages, distinct patients).
Technical replicates (duplicate wells, repeat reads) describe measurement precision and do not count toward n.
State n for every panel in the legend, with what one unit is ("n = 5 mice per group", "n = 3 independent experiments").
Pooling cells from many wells of one experiment and calling it n=many is pseudo-replication — a classic Cancer Cell reviewer catch.
Choosing the test
Design
Typical test
Two groups, continuous, ~normal
Unpaired t-test (Welch if unequal variance)
Two paired conditions
Paired t-test
Two groups, non-normal / small n
Mann-Whitney U
>2 groups, one factor
One-way ANOVA + post-hoc (Tukey/Dunnett)
Two factors (e.g., genotype × treatment)
Two-way ANOVA + correction
Tumor growth over time
Mixed-effects / repeated-measures ANOVA (not many t-tests per timepoint)
Survival / time-to-event
Kaplan-Meier + log-rank; Cox for covariates
Categorical / proportions
Fisher's exact / chi-square
Correlation
Pearson (normal) / Spearman (ranked)
High-dimensional omics
Model-based (DESeq2/edgeR/limma) with FDR
Check assumptions (normality, equal variance) and report how. Prefer non-parametric or Welch corrections for small/uneven bench-scale data.