Prepare cohort summaries
Separate unique participants, trips, and observations before drawing comparisons. Keep route, direction, and crew-role definitions explicit.
CASE STUDY 03 / RESEARCH TOOLS
Interactive aviation research dashboards
Turning complex research cohorts into explorable views without hiding denominators.
ROUTES / CREW / PHASESIllustrative visual; no research data shown
THE QUESTION
How can collaborators explore fatigue, sleep, and performance patterns while keeping the selected cohort visible?
THE SETTING
R Shiny notebooks prepare route-level source tables and expose filters for route, crew role, flight phase, operation type, and other research subgroups.
THE STORY IN PLAIN LANGUAGE
The interface makes the analytical reasoning available to other people.
Rather than sending a separate figure for every subgroup question, linked filters expose the cohort behind a view. Counts, plots, and tables give collaborators a way to investigate patterns while keeping the meaning of the denominator visible.
Technology behind this workflow 11 tools & formats
The stack follows the analytical task: prepare the data, fit or apply the method, check the result, and communicate the output.
See how the libraries connect to the wider portfolio →THE PROJECT, STEP BY STEP
See how the question became a reviewable result.
03 / RESEARCH TOOLSVISUAL WALKTHROUGH
STEP 01 / 03Prepare cohort summaries
Cohort definitionsReactive datasetCounts + charts
Routes / crew / phasesSTEP 02 / 03Build reactive exploration
Cohort definitionsReactive datasetCounts + charts
Routes / crew / phasesSTEP 03 / 03Preserve interpretation
Cohort definitionsReactive datasetCounts + charts
Routes / crew / phasesIllustrative process diagram
Build reactive exploration
Use Shiny filters to recalculate cohort counts, summaries, Plotly visualizations, and DT tables together.
Preserve interpretation
Present exploratory comparisons beside sample sizes and filtered records so viewers can understand what each chart includes.
THE TECHNICAL CASE
Decisions behind the analysis.
A research dashboard is an analytical interface. Every filter changes the population represented by the chart, so cohort counts and data tables need to update with the same logic as the visualization.
- Unit of analysis
- Participants, trips, and repeated measurements, each counted at its own level.
- Inputs
- Prepared route-level research tables with crew role, operation type, flight phase, sleep, fatigue, and performance measures.
- Deliverables
- Reactive Shiny views with Plotly figures, DT tables, and cohort summaries.
Prepare before presenting
The Shiny notebooks transform route-level source tables into fields suitable for filtering and display. This preparation includes operational categories and measurement phases so a filter has a defined meaning.
Reactive dependencies
Route, crew-role, flight-phase, and operation-type selections drive the research views. Shiny reactivity connects selection state to summaries, Plotly charts, and DT tables, reducing the chance that one view represents a different cohort.
Denominator-aware summaries
A unique participant, a trip, and a measurement row are different units. Counting rows as participants would overstate cohort size whenever a person contributes repeated observations. The display needs to keep these concepts explicit.
Exploration and interpretation
Interactive plots let collaborators examine subgroup patterns, while tables expose the records behind a view. Sample availability and sparse selections determine how much weight a visual difference deserves.
METHOD IN FOCUS
Selected cohort → shared reactive data → counts + summaries + plots + tables
A shared filtered dataset keeps the denominator consistent across outputs. An exploratory dashboard supports inspection and does not itself establish a confirmatory comparison.
EXPLORE THE METHOD
See a cohort change across connected views
A small synthetic dataset demonstrates linked filters, unique participant counts, repeat trips, and observed outcomes. This is a portfolio demonstration, not the internal Shiny application.
QUALITY OF THE EVIDENCE
Checks that change the conclusion.
- Empty selections
A filter combination with no eligible records should show an explicit empty state instead of an old chart.
- Consistent counts
Reconcile chart observations and table rows against the same filtered data and separately count distinct participants and trips.
- Missing outcomes
Show the number with an observed endpoint; do not let a total cohort count imply complete outcome availability.
- Filter combinations
Inspect route × role × phase combinations and avoid comparing groups with incompatible operational definitions.
FOR THE TECHNICAL CONVERSATION
Where I would take the discussion next.
Interview discussion can focus on reactive dependency design, the difference between input validation and statistical validation, and how a dashboard communicates uncertainty without overwhelming its users.
WHAT IT DEMONSTRATES
From method to meaning.
Researchers can inspect patterns by operational subgroup through interactive plots and tables.