Statistical consulting · Northampton, Massachusetts
Rigorous analysis built for impact.
Green Line Analytics provides statistical consulting, research support, and analytics instruction to institutions, researchers, and teams who need results they can defend, reproduce, and maintain after the engagement ends.
Services
Four practice areas, one standard of evidence.
Every engagement produces documented methodology, reproducible code, and results explained in language the audience actually uses — whether that audience is a dissertation committee, a board, or a team that has to run the analysis again next quarter.
Statistical consulting & quantitative analysis
Project-based analysis for institutions and organizations with data that doesn't fit a standard template.
- Study design, modeling strategy, and analysis plans
- Regression, multilevel models, survival analysis, causal inference
- Open-source implementation in Python and R — no license lock-in
- Documented, reproducible pipelines your team can maintain
Dissertation & research support
Methodological support for doctoral candidates and academic researchers. Statistical components only — analysis and interpretation, not data collection or writing on your behalf.
- Methodology consultation and analysis plan development
- Hypothesis testing, power analysis, advanced modeling
- Results sections prepared to publication and committee standards
- Preparation for the methods questions a committee will ask
Corporate analytics training
Hands-on workshops for teams updating their analytical toolkit, built around your data rather than generic exercises.
- Python and R for practitioners at any starting level
- Applied statistics, forecasting, and modern modeling methods
- Group, one-on-one, and multi-session program formats
- Delivered by a practicing analyst who teaches full time
Curriculum design
Analytics curriculum development for academic programs and corporate learning teams, aligned to what the field actually requires of graduates.
- Course and program design, from outcomes to assessment
- Alignment of academic content with industry practice
- Notebook-based instructional materials and autograded assessment
- Faculty and instructor enablement
About
Carey Baldwin
Twenty years of statistical analysis, consulting, and instruction — with a practice built on a single idea: an analysis is only finished when the client understands it well enough to defend it.
Carey teaches Python for business analytics at the undergraduate and graduate levels as a Senior Lecturer in Operations and Information Management at the Isenberg School of Management, UMass Amherst. Green Line Analytics is her independent consulting practice and operates separately from that role.
That combination is deliberate. Consulting work keeps the teaching current, and teaching enforces a standard of explanation that consulting rarely demands. Clients get both: analysis that holds up to scrutiny, and a clear account of why it was done that way.
- Experience20+ years in statistical analysis, consulting, and instruction
- MethodsClassical and modern statistical analysis, study design, modeling
- ToolsPython, R, SQL — open-source throughout
- AffiliationSenior Lecturer, Isenberg School of Management, UMass Amherst (listed for background; see note below)
Past and current work
Selected engagements and publications.
A cross-section of the practice: corporate training, curriculum development, peer-reviewed research in emergency medicine and higher education, and methodological support that carried through to publication.
Predictive analytics for a commercial banking team
A three-day applied program for bank analytics staff, built end to end on synthetic data modeled to match the structure of their own portfolio. Participants worked through logistic regression in Excel with Solver, Power BI dashboard construction, and model evaluation, supported by teaching guides written for later reuse.
The engagement closed with a complete deployment package — batch scoring script, HTML reporting dashboard, and audit artifacts — so the team could run, explain, and defend the models without further support. A mid-engagement review of the source data reframed the modeling approach at the customer level, with predictors drawn from deposit relationships and outcomes from the lending side.
Python for Data Science: a four-day intensive
A course developed and taught for an incoming cohort in MassMutual's Data Science Development Program. The curriculum moved from core Python programming and data structures through object-oriented design to the working data stack — NumPy and Pandas — and closed with configuring a productive development environment in Visual Studio Code.
Delivered remotely, pairing direct instruction with hands-on exercises, small projects, and an open Slack channel for discussion between sessions. The format was built so that new hires finished the week with a working environment and code they had written themselves, ready to contribute rather than ready to start learning.
Decrease in Trauma Admissions with COVID-19 Pandemic
A study of how pandemic social distancing affected surgical caseload. Trauma admissions and emergency general surgery cases were compared across five two-week periods between February and mid-April for each year from 2017 through 2020, using Poisson regression to model count data across time periods.
Trauma admissions fell 57.4% in 2020, with motor vehicle collisions dropping far more sharply than other mechanisms — 80.5% against 45.1%. Emergency general surgery volume showed no significant change, and no comparable shift appeared in any prior year, which isolated the effect to the pandemic period rather than to seasonality.
Kamine, T. H., Rembisz, A., Barron, R. J., Baldwin, C., & Kromer, M. (2020). Western Journal of Emergency Medicine, 21(4).doi.org/10.5811/westjem.2020.5.47780 →
Supportive Programs and Financial Aid: Measuring Their Impact on Retention of Blacks and Latinx College Students in the New England Region
Against a decade of declining enrollment in U.S. higher education, the study asked whether two institutional levers — the type of financial aid offered and the number of student support programs available — are associated with retention across New England institutions. The analysis paired t-tests with regression modeling to test both questions.
Need-based aid and the number of support programs both showed positive associations with retention for Black, Latinx, and White students in the region. The engagement began as dissertation methodology consultation and continued through to peer-reviewed publication and co-authorship.
Rodgers-Tonge, D., Wray, M., & Baldwin, C. (2023). Journal of Business Diversity, 23(4).doi.org/10.33423/jbd.v23i4.6614 →
Detecting unwanted tracking devices
An ongoing methodological project on a hard classification problem: deciding whether a device has been following a person, when the signal is rare, the background is dense, and a false alarm and a missed detection carry very different costs.
Work to date covers study design, construction of a labeled dataset with exact ground truth, and calibration of decision thresholds. Methods and results will be published when the work is complete. The engineering belongs to a separate venture; the methodology is the same work described above, applied to an unusually demanding problem.
Contact
Start with the question, not the method.
The most useful first conversation is usually about what you're trying to find out and what decision depends on it. Method comes after. No obligation, and no charge for working out whether this is a fit.
- Emailcbaldwin@greenlineanalytics.com
- LinkedInlinkedin.com/in/carey-baldwin
- LocationNorthampton, Massachusetts
- EngagementsRemote and on-site, New England