Quickstart

From your spreadsheet to defensible findings, in about five minutes.

DataStoryMD reads a real-world clinical cohort and writes up the associations that matter, adjusted and FDR-corrected, the way a biostatistician colleague would. This guide gets you from install to your first finding.

Setup

Get it running

Two ways in: try the demo to look around, or run it locally on your own data.

Just looking

Open the live demo

A synthetic 300-patient ovarian-cancer cohort, fully interactive. Nothing to install.

Open the demo →
Your own data

Run it locally

Runs entirely on your machine, fully offline. No terminal, no commands: you double-click to start. Currently in private beta.

How it works →

Local install (private beta)

Built so you never open a terminal:

  1. Install Docker Desktop (one free download from docker.com).
  2. Unzip the DataStoryMD folder you are given.
  3. Double-click Start-DataStoryMD. It sets up once, then opens DataStoryMD in your browser automatically.

No accounts, no commands. DataStoryMD makes no internet calls, so your data never leaves your machine. You can even run it with the network switched off.

Access is invite-only during the beta.

Your data never leaves your machine. No accounts, no cloud, no telemetry. Files are read in memory for the session and never written to disk or transmitted.
The workflow

Your first analysis

Upload a cohort (or click Try with sample dataset) and follow the tabs left to right.

1

Upload your cohort

Drag in an Excel or CSV file. Messy real-world exports are expected: coded columns, value legends in headers, mixed languages, sentinels.

.xlsx · .xls · .csv · .tsv
2

It maps your columns to clinical meaning

Headers are decoded, types inferred, values normalized. If your workbook has a Data Dictionary sheet, real names are recovered from it. Columns it can't confidently name are flagged, never guessed.

3

Discovery: read the findings

Results are grouped into clinical themes, ranked by impact. Every hazard ratio comes from an adjusted multivariable model; p-values are FDR-corrected. Biased comparisons (immortal-time, tautologies, surrogates) are excluded up front, each with a stated reason.

4

Explore, Overview, Pathway: interrogate it

Ask a question in plain English ("does complete debulking help platinum-resistant patients?"), stratify any outcome by any variable, or trace the causal map from baseline to outcome. Everything recomputes, adjustment and all.

5

Export for grants and talks

One click to a formatted Word or PowerPoint document: figures, methods, effect sizes, and citations included.

Reading the numbers

What a finding tells you

Every result carries its own methodology, so you can defend it.

BRCA2 germline → Overall Survivalsignificant
aHR 0.46  (95% CI 0.30–0.70)  ·  q = 0.002  ·  n = 300
aHR 0.46Adjusted hazard ratio: BRCA2 germline carriers have 54% lower mortality hazard, after adjustment.
95% CIThe plausible range for the true effect. It excludes 1.0, so the effect is statistically real.
q = 0.002The p-value after FDR correction for multiple testing, not a raw, un-adjusted p.
adjusted forAge, FIGO stage, histology, baseline CA-125: pre-specified confounders, via Cox proportional hazards.
n = 300Patients contributing to this model.
Good to know

Principles

DataStoryMD · clinical narrative intelligence · runs offline, on your machine