Capabilities
Independent checks for when a method or a data source changes
NakedSignal checks what happens to reported numbers when a method, instrument, data source, vendor or model changes. We take on a defined task at a fixed price, directly or as a subcontractor inside a prime contractor's existing contract.
What we do
Four deliverables
Method comparison and bridging
Source-change and data quality checks
Model and estimate verification
A methods note you can hand on
How we work
A defined task, fixed in advance
Fixed price per task
Runs where the data is
Your terms
Worked examples
Public worked examples on this site
The sample report, the change explorer and the guides run on open-access studies, cited with their licences on the public data page; MedEval-1 runs on a public image collection, cited on its own page; the change check and the calculators run in your browser on your own numbers.
Sample report
A change report on a real platform change, built from a public CD4 method comparison: bias at each decision limit, patients across each limit and a re-test band.
OpenChange explorer
Open-access method comparisons analysed the same way: comparison line, bias at decision limits, and results crossing each limit against repeat testing.
OpenChange check
Paired results checked in your browser, with the file never uploaded. The full report is free on a public example.
OpenJaffe or enzymatic creatinine
A method change worked through on public comparisons: what traceability fixes, what it does not, and how many results move across a line.
OpenMedEval-1
A versioned evaluation standard for medical AI, with a held-out track that is sealed, scored and rotated, demonstrated on public data.
OpenFree tools
Calculators for a reference change value, eGFR after a creatinine method change and the expected result on another analyser, each running in your browser.
Open
Status
Where we are today
Tested on public and synthetic data; not yet run on a customer's data.
We have no federal contract or subcontract history yet. We would welcome a small first task, scoped so that you can judge the fit from the result.
Talk to us about a task
Tell us what is changing, in which data, and what the result has to support.