From crim-skills
Guides analysis and reporting for Criminology manuscripts: count models, longitudinal panels, trajectory models, survival/recidivism, robustness checks, and reproducibility.
How this skill is triggered — by the user, by Claude, or both
Slash command
/crim-skills:crim-data-analysisThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
*Criminology* reviewers are methodologically sophisticated and increasingly expect that your results
Criminology reviewers are methodologically sophisticated and increasingly expect that your results
can be reproduced from deposited materials (see crim-data-and-transparency). Analyze as if both are
true. This skill covers execution and reporting norms; design decisions live in crim-research-design.
renv.lock, requirements.txt, recorded ssc/net/traj installs).Criminology reviewers are quantitatively literate and will name a mismatch between model and the data-generating process for offending. Use the outcome to pick the estimator, then defend the assumption the reviewer will probe.
| Outcome shape | Default estimator | Reviewer will probe |
|---|---|---|
| Over-dispersed offense counts | negative binomial w/ exposure offset | dispersion test, offset justification |
| Excess-zero counts (most offend zero times) | zero-inflated / hurdle | what the inflation stage means theoretically |
| Repeated within-person offending | fixed-effects / hybrid panel | within vs. between separation |
| Time-to-recidivism, censored | Cox / competing-risks | proportional hazards, censoring mechanism |
| Developmental offending paths | GBTM / growth mixture | BIC, AvePP ≥ 0.7, group shares not reified |
Suppose a hybrid model returns an incident-rate ratio of 0.62 on the within-person "employed" indicator (illustrative): entering employment maps to roughly a 38% lower offending rate for the same person, 95% CI [0.49, 0.78], net of stable traits. The between-person column, IRR 0.80, is weaker and reading it as desistance would conflate selection (people prone to desist also find work) with the within-person change the life-course claim requires. Report both; tell the reader which identifies the mechanism.
【Main estimate】magnitude (IRR / predicted count / hazard) + interval + substantive meaning
【Crime measure】reported / victimization / self-report + dark-figure caveat
【Within vs between】isolated correctly? [Y/N/NA]
【Model fit】counts: dispersion handled? trajectory: BIC/AvePP reported? [Y/N/NA]
【Robustness】specs that could break it → what held
【Reproducible】master script + seeds + pinned versions? [Y/N]
【Next】crim-tables-figures
../../resources/external_tools.md — count, trajectory, survival, and spatial packages../../resources/official-source-map.md — transparency expectations and crime-data sourcesnpx claudepluginhub brycewang-stanford/awesome-journal-skills --plugin crim-skillsDefends research designs for criminology manuscripts: causal identification, longitudinal/life-course methods, criminal-career trajectories, place-based experiments, and qualitative case selection. Mode-aware and tailored to journal standards.
Guides analysis and reporting for ASR manuscripts to meet masked review standards across quantitative, demographic, comparative-historical, and computational sociology.
Guides statistical analysis and APA reporting for communication research manuscripts, covering ANOVA, regression, SEM, mediation/moderation with honest uncertainty and reliability.