Skip to content
Data Reflector

Product

Statistics you can ask for.

Plain-English questions in, rigorous analysis out. A private engine does the computing where your data lives.

Start free

How it works

  1. 01 Load Your file, on your machine
  2. 02 Ask “Is age linked to readmission?”
  3. 03 Analyze Right test, checked by guardrails
  4. 04 Export Word, LaTeX, PowerPoint

What it does

Ask

Conversational, agentic analysis

Ask in plain English. The agent works out what you mean, runs the right operation, and explains the result. Follow-ups like “do it” continue where you left off.

Load

Your files, as they are

CSV, TSV, TXT, multi-sheet Excel and SAS XPT. Column types are inferred automatically, and you can adjust them.

Filter

Cohorts in plain English

Describe a filter or cohort and it is built for you, including filters that span related tables.

Explore

Exploratory analysis done for you

Summary statistics, distribution fitting and plots — per column, per table, or across every table at once.

Analyze

Trend and forecasting

Analytics tools that follow the same recommend-then-run pattern as the statistics.

Export

Ready for your paper or deck

Word, PDF, PowerPoint, LaTeX, Markdown, Excel, CSV, JSON, HTML, Parquet, SQLite or the clipboard. Interactive charts in the browser.

Audit

A record of every session

Each conversation and its results are logged, so analyses can be reviewed and reproduced.

Analyses

The right test. Not just a chart.

More than 30 analyses, including the standard UCLA statistics curriculum. Describe your question and a recommender picks the correct test.

Group comparisons

  • t-tests
  • ANOVA, factorial ANOVA, ANCOVA
  • Repeated measures
  • Nonparametric tests
  • Post-hoc tests

Association

  • Correlation
  • Chi-square and association
  • Agreement
  • Diagnostic accuracy

Regression

  • Linear
  • Logistic
  • Multinomial
  • Cox proportional hazards
  • Mixed models

Survival and more

  • Time-to-event analysis
  • Count models
  • Multivariate methods
  • Assumption checks

Guardrails

Warns or stops when the numbers would mislead.

Too few events per variable

Warns before a regression model is fitted on too little information.

Degenerate predictors

Catches constant or near-constant variables that make results meaningless.

Pseudo-replication

Flags repeated measurements treated as independent observations.

Start on the Free plan.

Sign up, install the engine with one command, ask your first question.

Start free Compare plans