Work with me.
I take on a small number of outside engagements alongside my main work, where the problem is interesting and the scope fits. This page describes how I usually work with collaborators, what I've shipped with others, and how to get in touch.
Two tracks, one set of methods.
- Co-authorship on papers in spatial statistics, Bayesian hierarchical modelling, healthcare cohorts.
- Methods consulting on model selection, validation strategy, panel-data design.
- Reproducible analysis pipelines for ongoing studies (R / Python / Stan).
- Statistical review for manuscripts and grant submissions.
- Power BI / Tableau dashboards end-to-end, from data model through layout and governance.
- KPI / funnel / cohort analysis for product or operations teams.
- Healthcare data analysis on panel, longitudinal, and claims-style data.
- Statistical model build with productisation hand-off.
Click a tag for where I've used it.
Methods, tools, and domains across the projects below. Tag size reflects how central it has been to my work. Click for the where-and-output breakdown.
What I've shipped with others.
KPI Metrics Establishment Committee
Cross-functional initiative to define a single canonical "active member" metric and ship the Power BI dashboard leadership reviewed weekly. I operated the committee and owned the data-and-dashboard build end to end; colleagues contributed prior-art review and analytical sanity-checking.
Healthcare cost prediction model (diabetes)
Random Effects Linear Regression on log-cost over five years of Korea Health Panel data, diabetes cohort (~1,400 patients/year). Variable selection across four derived specifications, validated on multicollinearity (VIF) and autocorrelation (Durbin–Watson), then productised as a member-facing lifestyle savings calculator.
Employee Assistance Program: clinical validation & B2B dashboards
Clinical validation of pre/post effects on a chronic-disease cohort, plus a workforce-level B2B dashboard for HR / benefits leaders. Three workforce views (Reach / Intensity / Outcomes) with row-level security per enterprise tenant. The dashboard structure also anchored the commercial pitch.
National Fitness 100, representativeness study
Ten-year exploratory study on Korea's national physical-fitness assessment dataset (2013–2022, 680K observations across 17 testing centres). Profiled demographic-distribution shifts year over year, decomposed missing-value structure across the adult and senior cohort splits, and triangulated against the Korea National Health and Nutrition Examination Survey (KNHANES) to assess representativeness for downstream policy modelling.
Handball player selection: early-stage talent ID
Machine-learning study to predict national-team selection from physical and psychological measurements, so the model could surface promising players before a long match history accumulates. Four years of data (2019–2022), roughly 1,900 players split into field and goalkeeper sub-models. Pipeline: MICE imputation, Multiple Linear Regression on FILA point, Factor Analysis (KMO 0.84) with Cronbach's alpha validation, LASSO for multicollinearity, PCA, then Random Forest classification with SMOTE for the severe class imbalance (72 selected vs 1,517 not).
Spatial Two-Stage Hierarchical Optimisation for Bitcoin Mining
Two-stage spatial framework for site selection of Bitcoin mining facilities on surplus solar electricity in South Korea. I led methodology, code, and writing across two preprint iterations.
Nonparametric bootstrapping for spatial point patterns
Summer research internship: bootstrap inference for spatial point processes with measurement error, and computationally efficient methods for very large datasets. Concluded with a poster presentation.
Graduate research, ongoing
MS thesis in progress. Mini-thesis already shipped: a Bayesian hierarchical multivariate spatial model on the prevalence of diabetes and hypertension across Korean administrative districts, joint with national-level survey data.
From inquiry to hand-off.
Email with the prompts in the next section, or a brief description if simpler.
Fit, scope, constraints, timing. Free, no commitment either way.
Deliverables, milestones, fee (if commercial), authorship and credit terms (if academic).
Weekly check-ins. Code, write-up, and presentation deliverable at close.
What to send in your first email.
Email me at clodaghchoi@gmail.com. A brief paragraph is fine; a short structured note is faster for both of us.
- // Context
- Academic collaboration, commercial project, or personal initiative.
- // Stage
- Early ideation, mid-engagement, or wrapping up. Either is fine.
- // Scope
- What you have already (data, draft, notebooks, dashboard mockup) and what you'd like from me.
- // Timeline
- Rough is fine. Flexible engagements work; hard deadlines are workable if surfaced early.
- // Compensation expectation
- Only if commercial. Academic collaborations typically run on authorship credit and reasonable expense reimbursement.
I'll respond to every inquiry, but I take on a small number of engagements at a time. If the timing or scope doesn't fit on my end I'll say so quickly, and where I can I'll suggest someone better placed.