radiology-stats
Plan/compute/audit inference on defined data/effect rows: CI, AUC, MRMC, survival, meta-analysis and power. CN: 统计、样本量、AUC比较、校准
Install / Use
npx skills add huang-sir1/radiology-skills --skill radiology-statsInstalls into whichever agent you are using.
SKILL.md
Installable skill definition
Quality Score
Category
Development & EngineeringSupported Platforms
Tags
Our assessment of radiology-stats
radiology-stats scores 87/100 on our quality scale, 2165th of 4,700 Development & Engineering skills we index (top 47%).
Its SKILL.md is 17 KB long, well organised into 9 sections and no code examples: a thorough specification that gives an agent plenty to work with.
With 1,785 GitHub stars, it is one of the more widely adopted skills in the catalogue.
Maintenance, license and trust
- The repository was last updated about 2 months ago, so radiology-stats is actively maintained.
- Our last check on 2026-11-02 found the source still online.
- No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
- Its trust signals score 88/100, with 1 caution from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.
AI review by kimi-k2.7-code on 2026-10-02. Automated pattern scan on 2026-10-02. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
radiology-stats compared with similar skills
All 4 of these similar skills score higher than radiology-stats; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| radiology-stats (this skill)by huang-sir1 | 87 | 1.8k | 56d ago | SKILL.md |
| taste-skillby Leonxlnx | 100 | 116.8k | 8d ago | CLAUDE.md |
| ai-job-searchby ai-job-search-dev | 100 | 72.1k | today | CLAUDE.md |
| claude-howtoby luongnv89 | 100 | 47.3k | 2d ago | CLAUDE.md |
| lobehubby lobehub | 100 | 85.6k | today | MCP Server |
Frequently asked questions
- How do I install radiology-stats?
- Run
npx skills add huang-sir1/radiology-skills --skill radiology-stats. The install tabs above show the steps for each supported agent. - Which AI agents does radiology-stats work with?
- It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
- Is radiology-stats safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It declares no license and scores 88/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is radiology-stats still maintained?
- The repository was last updated about 2 months ago, so radiology-stats is actively maintained.
Skill content
View source on GitHubname: radiology-stats description: "Plan/compute/audit inference on defined data/effect rows: CI, AUC, MRMC, survival, meta-analysis and power. CN: 统计、样本量、AUC比较、校准"
Imaging Biostatistics for Radiology
Use this skill to choose the right test, run it correctly, and report it the way Radiology wants — estimates with 95% CIs, exact p-values, named tests, and multiplicity handled. It covers the statistics that imaging-AI, radiomics, and reader studies live or die on.
Core stance
- Estimate + uncertainty, not just p. Every primary result gets a 95% CI. Report exact
p-values (e.g.
P = .03, notP < .05); useP < .001only below that floor. - The test must match the design. Paired data → paired test (same patients/cases read by both methods); clustered data (multiple lesions per patient) → account for clustering; multiple readers → MRMC, not a naive average.
- Match performance measures to use. A probabilistic diagnostic/prognostic model generally needs discrimination and calibration. Evaluate decision-analytic utility only when an intended decision, threshold range, consequences and comparator strategies are defined; do not require a decision curve for every exploratory classifier.
- Control multiplicity honestly. Thousands of radiomic/omic features ⇒ FDR or stronger; pre-specify primary vs exploratory.
- No fishing, no fabrication. Pre-specify the primary analysis; never invent a number, a CI, or a p-value. If data are insufficient, say what is needed.
- Reproducible. Always return an executable statistical brief. Return runnable code and computed results only when the required data/schema, estimand, hierarchy, environment and execution authority are available; record software/version and the exact uncertainty method.
- Evidence state is explicit. Use
PLAN_ONLY,DATA_READY,COMPUTED,AUTHOR_REPORTED,PARTLY_VERIFIED,NOT_ASSESSABLE, orBIOSTATISTICIAN_REQUIRED. A proposed model is not a run. - No universal sample-size threshold. Power and precision depend on the estimand, endpoint distribution, effect worth detecting, allocation unit, clustering/repeated structure, multiplicity, attrition and acceptable uncertainty. Do not substitute an EPV, event-count or replicate rule of thumb for those inputs.
Choose the route
| Route | Use when | Required contract |
|---|---|---|
| single-study-inference | estimating or comparing outcomes within one imaging/omics cohort or reader study | endpoint, estimand, true unit, pairing/clustering, missingness, model/test, uncertainty and multiplicity |
| meta-analysis-inference | compatible study/effect rows need a synthesis model, audit or result interpretation | study-family/effect IDs, effect scale/direction, unit and dependence, model family, heterogeneity, sensitivity and small-study boundary |
| experimental-inference-power | a tissue, cell, organoid or animal experiment needs inferential design, model or power/precision planning | assigned unit, endpoint distribution, estimand, cluster/repeat hierarchy, effect worth detecting, nuisance inputs, attrition and feasible design |
| quantitative-measurement | repeatability, reproducibility, bias, change detection or scanner/site comparability of a quantitative image measure | measurand, repeat/reproduce conditions, bias reference, variance components, wCV/repeatability coefficient/LoA or smallest detectable change and claim boundary |
| causal-clinical-impact | estimating workflow/intervention effects from randomized, pragmatic or observational deployment data | target trial/contrast, assignment/exposure time, confounding, interference, missingness and sensitivity plan |
| longitudinal-multistate | repeated imaging, landmark/dynamic prediction, interval-censored transitions or competing events | time zero, visit process, time-varying exposure/measurement, state definition, estimand and dependence model |
Declare one mode:
| Mode | Output boundary |
|---|---|
| plan | executable brief and missing-input ledger; no computed result |
| audit | artifact-anchored design/model/inference findings, repairs and stop gate |
| compute | authorized code/run record and verified outputs from sufficient inputs |
| interpret | effect/uncertainty/heterogeneity/power meaning and bounded claim from real outputs |
| mentor | minimum defensible, standard publishable and stronger route with why, prerequisites, resource cost, failure signal, fallback, closure evidence and surviving claim |
| writing-handoff | fixed Methods/Results/table/figure/statistical-reporting fields without evidence upgrade |
If the request crosses routes, keep separate estimands and briefs. Open only the active specialized reference: meta-analysis, experimental inference/power, quantitative measurement, causal clinical impact, or longitudinal/multistate imaging. Persist the meta-analysis or experimental specification with the meta-analysis brief or experimental inference/power brief.
When to use
- Diagnostic accuracy: sensitivity/specificity/PPV/NPV/accuracy/likelihood ratios + CIs.
- Comparing models/readers/tests: DeLong or bootstrap for AUCs; McNemar for paired sensitivity/specificity.
- Reader studies: kappa / weighted kappa / Fleiss / ICC / Bland-Altman; MRMC analysis and
sample sizing (clinical reader-study design stays with
radiology-translation). - Prediction models: ROC, calibration (slope/intercept, Brier), decision-curve analysis, threshold selection.
- Radiomics/omics: feature reproducibility (ICC), multiple-testing correction, cross-validation/nested CV, bootstrap optimism.
- Statistical inference for already-defined configuration, method, ablation or sensitivity
comparisons: paired deltas, cluster bootstrap, run variability and multiplicity. The scientific
comparison design itself remains with
radiology-method-evaluation. - Survival/prognosis: Kaplan-Meier + log-rank, Cox, C-index, time-dependent ROC, competing risks.
- Quantitative imaging measurement: bias, repeatability/reproducibility, wCV, repeatability coefficient, limits of agreement, smallest detectable change and scanner/site variance.
- Causal/clinical impact: randomized/pragmatic, cluster/stepped-wedge, ITS/DiD or target-trial analyses with explicit identification assumptions and interference/time-varying confounding.
- Longitudinal imaging: mixed/joint/landmark/multistate models, interval censoring, informative visit/dropout processes and competing events.
- Planning: sample size/power/precision for accuracy, AUC, prediction models and experiments, using input-driven methods rather than universal EPV, event-count or replicate cutoffs.
- 中文高频说法:“样本量/统计功效怎么算?”“两个 AUC 怎么比较(DeLong 还是 bootstrap)?” “校准曲线/决策曲线怎么做?”“多阅片者一致性(kappa/ICC/MRMC)怎么报?”“meta 分析怎么合并?”
When to open extra files
| File | Open when | |---|---| | references/diagnostic-accuracy.md | Sensitivity/specificity/PPV/NPV/LR, the right CI method, paired comparison (McNemar) | | references/model-evaluation.md | ROC/AUC, DeLong vs bootstrap, thresholds, calibration, Brier, decision-curve analysis | | references/incremental-value.md | Added information versus validated predictive gain: distinguish development LRT, independent binary ROC and censored survival comparison; condition DCA and optional NRI/IDI on their decision and inference assumptions | | references/agreement-mrmc.md | Cohen/weighted/Fleiss kappa, ICC model choice, Bland-Altman, MRMC (Obuchowski-Rockette / DBM) | | references/high-dimensional-omics.md | Multiple testing (Bonferroni/Holm/BH-FDR/q-values), CV/nested CV, leakage, optimism, harmonisation stats | | references/survival-prognostic.md | Kaplan-Meier, Cox PH (+ assumptions), C-index, time-dependent AUC, competing risks | | references/sample-size.md | Input-driven sample-size/power/precision for sensitivity/specificity, AUC, prediction validation, MRMC and clustered designs without universal cutoffs | | references/analysis-hierarchy-and-missingness.md | Multiple lesions/slices/timepoints/readers, nested or repeated data, biological vs technical replication, missing data, exclusions, and analysis-plan deviations | | references/method-comparison-and-sensitivity.md | Real benchmark, ablation or parameter-sensitivity outputs need paired deltas, patient/donor-level uncertainty, run variability and multiplicity without treating folds/seeds as subjects | | references/meta-analysis-inference.md | Generic/dependent random-effects synthesis, bivariate/HSROC DTA, prevalence/incidence, HR/OR/RR, ICC/kappa/agreement, prediction performance/calibration, heterogeneity, few-study and reporting-bias boundaries | | references/experimental-inference-and-power.md | Continuous/binary/count experimental endpoints, donor/litter/cage/cluster assignment, repeated measures/mixed models, attrition and input-driven power/precision planning | | references/quantitative-imaging-measurement-science.md | Bias, repeatability/reproducibility conditions, variance components, wCV, repeatability coefficient, Bland–Altman limits, smallest detectable change and cross-site quantitative claims | | references/causal-and-clinical-impact-inference.md | Target-trial framing, randomized/pragmatic/cluster/stepped-wedge/ITS/DiD routes, confounding, interference and causal claim ceilings | | references/longitudinal-and-multistate-imaging.md | Repeated imaging, time-varying/landmark/joint/multistate models, interval censoring, visit-process bias and competing events |
Workflow
- Declare route, evidence and execution state. Inventory the supplied protocol, effect rows,
raw/summary data, code/output and authority. Never turn a brief or author-reported result into
COMPUTED. - Restate the design and hierarchy — independent experimental unit, patient/lesion/slice/
timepoint nesting, paired vs unpaired, repeated measures, readers, prevalence, primary vs
secondary endpoints. Open
analysis-hierarchy-and-missingness.mdwhen clustering or missingness exists; do not count images/cells/model runs as independent patients. - Pick the estimand and model/test using the route reference. Name the population quantity, contrast, endpoint, time/measurement period and intercurrent/missing-data handling explicitly.
- Choose the uncertainty method (e.g. Wilson/Clopper-Pearson for single-study proportions; DeLong or bootstrap for AUC; bootstrap for derived metrics).
- Handle dependence, missingness, exclusions, attrition and deviations before looking at significance; define the analysis population and sensitivity analyses.
- Handle multiplicity — declare the primary analysis; correct the rest (method + family).
- Run the stop gate. If an estimand, independent/allocation unit, dependence structure, effect
definition or required power input cannot be made defensible, return
BIOSTATISTICIAN_REQUIREDwith the exact blocking decision and do not present a final m
Truncated for display — read the full file on GitHub.
Related Skills
taste-skill
116.8kAgent skills that give coding agents taste: art direction, reference images and a browser check instead of generic AI pages
ai-job-search
72.1kThe job search that runs on your machine. AI job application framework built on Claude Code: evaluate postings, tailor CVs, write cover letters, prep interviews. Fork it and own it.
claude-howto
47.3kA visual, example-driven guide to Claude Code — from basic concepts to advanced agents, with copy-paste templates that bring immediate value.
lobehub
85.6k🤯 LobeHub - Super-Agents on Your Team. Add super-agents to the chats your team already uses.They take on work by themselves and get better every week.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
