Keep the starting
point in sight.
A fixed baseline preserves the original comparison. A rolling view shows how your pattern evolves. Explore both without quietly moving the goalposts.
Explore the two views BASELINE-ANCHORED SENSING & EVALUATION
Change is personal.
So is the evidence.
Understand what changed for you.
How much. How confidently. In what context.
Turn longitudinal wearable data into an inspectable record of personal change, anchored to where you started.
Working research preview · Synthetic demo · Local CSV analysis
ONE PERSON. TWO WINDOWS.ILLUSTRATIVE
Your reference
stays in view.
stays in view.
Evidence, attached.Reference · uncertainty · context
You, over time.
Population knowledge provides context. Your own history provides a personal reference. BASE makes that comparison visible and reviewable.
Keep the starting
Ask what else
could explain it.
Travel. Illness. Missing data. Inspect the context and compare results with flagged days excluded. The uncertainty belongs beside the estimate.
Test a confounded scenarioTake the reasoning
with the result.
Export the windows, methods, coverage, assumptions, and data fingerprint behind your analysis. Every result should be possible to inspect.
Create an evidence receiptTHE FIRST QUESTION IS YOURSDid the change
Did the change
change anything?
Explore a sleep schedule, inspect a travel confound, or bring a daily CSV from one individual. Start with a question you can actually measure.
Open the working preview 3 scenarios · 3 outcomes · Transparent methods
Built for exploration and pilot design. Clinical validation, direct wearable connections, and team workflows are future development.
OBSERVE · COMPARE · UNDERSTAND
Your evidence studio.
One person. A clear question. An inspectable answer.
THE QUESTION
After a protected sleep schedule begins, how does sleep duration change from this person’s baseline?
THE PERSONAL TRAJECTORY
Sleep duration over time
Daily observation Fixed baseline mean Trailing 7-day mean Baseline middle 80%
Shaded range describes baseline observations; it is not a clinical reference interval.● Context flagged × Not usable
Inspect daily records
| Date | Phase | Value | Use | Context |
|---|
WHAT THE EVIDENCE SUPPORTSOBSERVATIONAL
An analyst-selected absolute magnitude, not a validated clinical cutoff. Changes to this setting make this an exploratory analysis.
01 / Observed change02 / Adjusted association03 / Causal effect
This release estimates observed change. It does not establish intervention efficacy or readiness for duty.
CONTEXT SENSITIVITY
Does the story hold?
Compare all usable days with a view that excludes context-flagged days.
This comparison is descriptive. Excluding days does not adjust for confounding and can introduce selection bias. Inspect why each day was flagged.
DATA SUFFICIENCY
Research exploration only. Personal history complements clinical context and established safety thresholds. Read the method and limitations
OPEN METHODS / VERSION 0.1
Show the work.
Keep the uncertainty.
BASE separates what was observed from what can be concluded. The first release is a descriptive, within-person pre/post analytics tool.
Two views, distinct jobs.
The fixed baseline is the mean of usable daily values before the event, inside your selected baseline window. The chart’s shaded band spans the empirical 10th to 90th percentiles of those observations.
The adaptive view is a trailing seven-calendar-day mean, shown only when at least four usable days are present. It can cross the event boundary and is a descriptive monitoring view. It never replaces the fixed reference in the pre/post estimate.
Change, with an interval.
The estimate is the follow-up mean minus the baseline mean in the original outcome units. BASE uses 1,200 deterministic moving-block bootstrap draws, resampling each phase independently, to produce an approximate 95% percentile interval.
Blocks contain consecutive calendar days and never cross gaps, phase boundaries, or excluded days. Select 2, 3, or 7 days to explore sensitivity to short-term dependence. A fixed seed makes the same analysis reproducible.
The bootstrap assumes approximately stationary observations within each phase. It does not remove time trends, seasonality, regression to the mean, or unmeasured confounding. Block length and short windows can materially affect coverage. This engine has not undergone clinical validation.
Pre/post is not proof of cause.
Timing alone cannot attribute a change to an intervention. Formal N-of-1 trials use stronger designs, often repeated randomized comparisons with attention to carryover. This release does not implement randomization, adjusted causal models, or trial registration.
A threshold crossing indicates change relative to an analyst-selected magnitude. It is not automatically beneficial, clinically meaningful, or evidence of operational readiness. Multiple outcomes and repeated window exploration increase the risk of misleading findings.
An exportable evidence receipt.
The JSON receipt records the dataset’s SHA-256 fingerprint, analysis windows, outcome, thresholds, exclusions, coverage, method version, estimates, and limitations. The printable HTML report makes the same result easy to review.
Receipts are unsigned analysis snapshots. They do not certify provenance, prove data authenticity, or establish prospective registration. Raw daily records are excluded from the receipt; save your source data separately for reproduction.
Browser-local by design.
CSV data is parsed and analyzed in this browser’s memory. No record upload, third-party analytics, advertising pixels, local storage, or backend account is used. Refresh or close the page to clear the imported dataset. Downloads remain wherever you save them.
The hosting provider processes ordinary page requests. Direct wearable API connections, cloud storage, accounts, team access controls, and clinical workflows are not included in this preview.
METHOD REFERENCESSoltani et al. (2025) · Intensive longitudinal sampling and within-person variability AHRQ · Design and Implementation of N-of-1 Trials: A User’s Guide CENT 2015 · Reporting N-of-1 trials
Explore the evidence studio These references inform the design. They do not validate or endorse BASE.