Many questions.
One analytical thread.

Analyzed how cognitive performance returns to a pre-trip baseline after long-haul flights. The work links pre-flight and post-flight Psychomotor Vigilance Task (PVT) measurements with sleep, layover, and demographic variables. Recovery notebooks construct event indicators and follow-up times, fit Cox models, and evaluate separate 24-, 48-, and 72-hour sleep predictors. Earlier notebooks explore repeated-measures models and Poisson alternatives.

Methods & decisions

Cox proportional hazards; censoring; longitudinal analysis; linear mixed-effects models; Poisson model evaluation; residual diagnostics; feature engineering

Libraries & tools

Why it transfers

Time-to-event research where follow-up is incomplete and endpoint definitions matter.

Integrated wearable actigraphy exports with flight records and sleep logbook information to derive analysis-ready sleep variables. The scripts parse heterogeneous exports, align sleep episodes to trips, distinguish total sleep time from time spent in a sleep episode, and calculate sleep inside windows relative to flight events. This work demonstrates the data engineering behind the statistical analyses: reconciling formats, maintaining trip keys, handling midnight boundaries, and delivering reusable features to downstream models.

Methods & decisions

Data integration; temporal joins; interval overlap; timestamp normalization; repeated-trip alignment; derived features; missing-data handling

Libraries & tools

Why it transfers

Longitudinal sensor integration, temporal joins, feature preparation, and scientific ETL.

Examined how layover duration, arrival timing, direction of travel, crew role, and inflight sleep relate to sleep during layovers. The workflow assembles route-level data, anchors timestamps to home-base and destination time, calculates sleep quantity and quality measures, and compares shorter and longer layovers. Model scripts include participant random effects and distribution-specific models for outcomes such as wake after sleep onset and sleep latency. Visualization and manuscript-oriented code provide tables, timelines, and route comparisons.

Methods & decisions

Mixed-effects regression; Gamma generalized mixed models; repeated observations; exploratory analysis; circadian alignment; sleep efficiency

Libraries & tools

Why it transfers

Repeated-measures research with participant effects and distribution-specific outcomes.

This work includes shared and legacy analysis code; the portfolio describes contributions to that workflow.

Compared operational rest schemes and pre-flight nap groups using cognitive performance, sleepiness, fatigue, and sleep measures. The work classifies rest opportunities into first, middle, second, and split-break patterns; separates flying and relief crew; and examines measurements at flight and break timepoints. Report notebooks produce subgroup counts, time profiles, comparisons, and effect sizes.

Methods & decisions

Group comparisons; Welch t-tests; Cohen’s d; confidence intervals; subgroup analysis; complete-case counts; repeated-timepoint visualization

Libraries & tools

Why it transfers

Observational comparison, segmentation, and communication of uncertainty.

Created visualizations showing how pre-flight naps and inflight sleep align with flight timing and the Window of Circadian Low (WOCL). The plots place individual trips on a shared time axis, overlay circadian windows, and color sleep segments by break schedule. Companion notebooks compare safety performance indicators over flight timepoints and provide interactive views.

Methods & decisions

Time-series visualization; interval alignment; home-base-time normalization; Window of Circadian Low overlap; subgroup exploration

Libraries & tools

Why it transfers

Event-stream visualization, interval alignment, and scientific data storytelling.

Built interactive research interfaces for exploring fatigue and sleep measurements by route, crew, direction, flight phase, and operation type. Larger notebooks combine source-data preparation with reactive filters, plots, participant/trip counts, and data tables. Variants address AMOC versus SSO operations, W/tag trips, time-of-day groups, long-range versus ultra-long-range flights, and flying versus relief crew.

Methods & decisions

Reactive programming; interactive filters; data aggregation; cohort comparison; exploratory interfaces; interactive tables

Libraries & tools

Why it transfers

Interactive analytics products with consistent filter logic and clear denominators.

Prepared route-comparison datasets and report outputs for long-haul aviation fatigue studies. The common workflow derives analysis variables, chooses usable trips, assembles paired or matched cohorts, compares sleep and performance measures, and produces reusable visual panels and statistical appendices. Variants support LAX–MEL/LAX–PVG, SFO–MEL/comparator routes, IAH–SYD, and SFO–MNL PM.

Methods & decisions

Paired comparisons; t-tests; TOST equivalence testing; non-inferiority confidence-interval analysis; parameterized reporting; descriptive statistics

Libraries & tools

Why it transfers

Statistical programming, comparative analysis, and parameterized reporting.

Comparative reports support research interpretation; they do not establish regulatory approval or causal safety effects.

Constructed comparable flight-trip cohorts for statistical route comparisons. Main scripts read prepared datasets, retain eligible outbound/inbound records, call a best-match function, and write matched-pair workbooks. The matching code evaluates operational differences such as duty time at top of descent, time awake, flight duration, and flight duty period, with role-based restrictions. Separate variable-preparation scripts derive sleep and schedule features from the master database.

Methods & decisions

Custom matching algorithms; eligibility filters; record linkage; missing-value handling; feature derivation; modular functions

Libraries & tools

Why it transfers

Record linkage, cohort assembly, custom matching logic, and reusable data preparation.

The implementation uses operational rule/distance matching; it is not presented as propensity-score matching.

Estimated recruitment requirements for fatigue and performance comparisons using existing route data. The notebooks derive outcome variability and observed group differences, apply analytical non-inferiority sample-size functions, and check assumed designs through simulated studies. The implementations distinguish paired and unpaired analyses and use different margins for PVT and subjective scales. Route-specific variants support California/Melbourne, Adelaide planning, SFO–MNL, SFO–DEL, Island Hopper, and other operational studies.

Methods & decisions

Power analysis; non-inferiority study design; paired/unpaired designs; Monte Carlo simulation; effect-size assumptions; sensitivity analysis

Libraries & tools

Why it transfers

Experiment planning, non-inferiority design, and sensitivity analysis.

Built simulation functions to explore the feasibility of repeated-measures fatigue studies before final data collection. The scripts generate participant covariates, repeated trips and timepoints, sleep predictors, random participant effects, and configurable missingness. They fit mixed-effects models to simulated continuous or binary outcomes and summarize the frequency of detecting selected effects. A separate version focuses on four flight legs without missingness. This demonstrates statistical programming and model-evaluation design, with a clear distinction between simulated behavior and observed operational results.

Methods & decisions

Synthetic data generation; Monte Carlo power; linear and logistic mixed models; random effects; intraclass correlation; missingness scenarios; reproducible random seeds

Libraries & tools

Why it transfers

Simulation-based evaluation of clustering, missingness, and design choices.

Simulation results describe assumed data-generating scenarios rather than completed-study findings.

Developed analysis specifications and exploratory outputs for multi-segment W-pattern and tag-flight operations. The plans define a primary second-leg top-of-descent comparison, secondary longitudinal questions, fatigue endpoints, sleep and recovery measures, and reporting requirements. Visualization notebooks inspect route patterns and performance across segments. This project demonstrates translating stakeholder research questions into implementable analyses and aligning cohort definitions with complex trip structures.

Methods & decisions

Statistical analysis planning; endpoint definition; cohort design; longitudinal visualization; multi-segment trip analysis

Libraries & tools

Why it transfers

Translating a new research question into a statistical analysis plan and data specification.

Planning and exploratory work; the proposed analyses are not presented as completed study results.

Automated the preparation of pilot-data collection metrics from master tracking and existing summary workbooks. An orchestration script reads source tracking data, calculates complete/incomplete statuses, updates route-by-month metrics and annual summaries, and writes a formatted workbook. Route/program-specific functions handle different operational record structures.

Methods & decisions

Batch processing; workflow orchestration; data reconciliation; completion-status rules; route/month aggregation; formatted workbook generation

Libraries & tools

Why it transfers

Reporting automation, deterministic transformations, and metric reconciliation.

Created views of how much usable data is available for each route, crew, timepoint, and outcome. The scripts count pre/post PVT observations, identify participants meeting baseline and follow-up requirements, track repeat studies, and summarize operational measures for monthly reports. Route-specific sample reports make missingness and eligibility visible before statistical modeling.

Methods & decisions

Data completeness profiling; cohort eligibility; distinct participant counts; repeated-trip tracking; descriptive reporting; reporting automation

Libraries & tools

Why it transfers

Data-quality assessment, cohort feasibility, and transparent reporting of missingness.

Implemented a workflow that evaluates flight records against configured duty-limit rules and produces flagged-record lists and audit workbooks. The code handles time zones, crew-size branches, trip segments, acclimatization assumptions, and cases needing review. A separate validation script recomputes expected fields and compares them with generated outputs. This provides a concrete engineering story involving deterministic business rules, traceable transformations, and verification of saved results.

Methods & decisions

Rule-based classification; time-zone conversion; boundary conditions; source-to-output reconciliation; audit trails; duplicate-key checks; exception reporting

Libraries & tools

Why it transfers

Business-rule implementation, boundary handling, exception reporting, and auditability.

Prepared a multi-route analysis of flight operation type, sleep timing, and cognitive outcomes. The code reshapes outbound/inbound records into trip-level data, derives operational timing and WOCL-overlap variables, and generates subgroup visualizations. Diagnostic sections compare raw and log-transformed reaction-time models and a Gamma mixed-model alternative. Accompanying documents record the team’s evolving decisions on primary comparisons and future outcome-specific models.

Methods & decisions

Trip-level reshaping; circadian-overlap features; random-intercept models; log transformation; Gamma GLMM; residual diagnostics; sensitivity modeling

Libraries & tools

Why it transfers

Distribution-aware modeling, residual investigation, and model sensitivity.

Exploratory model diagnostics; the final combined-study conclusions are outside this summary.

Examined how departure timing relates to cognitive performance, fatigue, and sleepiness in Island Hopper operations. The notebook fits separate outbound and inbound regressions and translates hourly coefficients into an estimated change for a 30-minute schedule shift. Related calculations explore how adjusting for departure timing affects outcome variability and planned sample sizes.

Methods & decisions

Linear regression; schedule sensitivity; effect-size interpretation; covariate-adjusted variance; sample-size planning

Libraries & tools

Why it transfers

Translating regression coefficients into understandable scenarios without overstating causality.

The schedule-shift summaries describe associations, not causal effects of changing a schedule.

Explored fatigue and sleepiness differences by nap status, flight direction, and measurement phase in the UWON dataset. The notebook reshapes survey outcomes, builds report-time and top-of-descent plots, and compares nap groups and within-participant changes. A mixed-model section accounts for participant clustering and includes residual, influence, and collinearity diagnostics.

Methods & decisions

Welch t-tests; paired comparisons; participant random effects; residual checks; influence diagnostics; subgroup visualization

Libraries & tools

Why it transfers

Clustered-outcome analysis, subgroup exploration, and influential-observation assessment.

Exploratory survey models; final validated nap-effect estimates are not claimed.

Prepared descriptive reports of pilot responses before and after focus-group activities related to sleep, fatigue, and rest guidance. The script imports separate survey exports, selects yes/no items, reshapes questions for aggregation, applies readable question labels, and plots response distributions in questionnaire order.

Methods & decisions

Survey recoding; categorical summaries; long-format reshaping; questionnaire labeling; pre/post descriptive reporting

Libraries & tools

Why it transfers

Survey analytics, categorical transformations, and stakeholder-facing visual reporting.

Descriptive pre/post reporting; the work does not establish intervention effectiveness.

Summarized pilot commuting patterns and whether residence and home base share a time zone. The notebook reads questionnaire data, applies usable-record filters, and calculates counts and percentages for commuting and time-zone groups. Base-specific summaries provide operational context for research on pre-trip sleep and fatigue.

Methods & decisions

Cohort segmentation; categorical proportions; data cleaning; time-zone grouping; descriptive analytics

Libraries & tools

Why it transfers

Segmentation, questionnaire data cleaning, and descriptive population profiling.

Time-zone categories do not measure commuting distance or travel time.

Maintained supporting assets that connect the wider research program. Data-preparation notebooks standardize source fields, derive flight and sleep measures, and export reusable workbooks. A layover calculator produces interval-based variables for downstream analyses. Documentation records assumptions, validation checks, and reporting steps for research handoffs.

Methods & decisions

Dataset standardization; date/time conversion; reusable calculations; analytical documentation

Libraries & tools

Why it transfers

Modular transformations, shared analytical definitions, and reproducible handoffs.

Shared workflow documentation and data preparation; no LLM application is implied.