How it works
KyroStat walks you through the same workflow a statistician would follow: inspect the data, choose a justified test, run it on real engines, and report it properly. Here is what happens at each step.
Start with the file you already have: a CSV or Excel spreadsheet exported from your survey tool, lab instrument, or data collection sheet. There is no import wizard to configure and no fixed-width format to wrestle with.
KyroStat reads your columns and detects variable types automatically, distinguishing the numeric measures you will analyse from the categorical variables you will group by. You can review and correct the detected types before anything else happens.
Before you run a single test, KyroStat computes descriptive statistics for every variable: means, standard deviations, counts, and distributions. This is the same discipline a statistician would insist on, made automatic.
Reviewing descriptives first catches data problems early: a column read as text, an impossible value, a group with three participants. Five minutes here saves a retraction-grade mistake later.
Write your research question in plain language: "Does the intervention group score higher on the post-test than the control group?" KyroStat reads the question together with your dataset and recommends between 3 and 8 statistical tests and 2 to 5 plots.
Every recommendation comes with a one-sentence rationale tied to your specific question, never a generic description of what the test does. The AI is explicitly forbidden from recommending a test just because your data happens to contain numbers. You review the rationale and decide what runs.
When you run a test, KyroStat generates analysis code and executes it server-side in a sandbox running real Python and R statistical engines: the same scipy, statsmodels, and R routines a statistician would use. Nothing is approximated in the browser, and the AI never computes a statistic itself.
The generated code is part of your result. You can read exactly what ran, download it, and rerun it anywhere, which makes your analysis transparent and reproducible from day one.
Results arrive as SPSS-style tables for checking your work and APA 7 tables for the manuscript, alongside publication-ready plots. The exact APA result sentence is assembled deterministically from the computed statistics, down to the leading-zero conventions.
One click bundles everything into a formatted PDF report: tables, plots, the generated code, and a paste-ready Methods paragraph naming each test, engine, and package version used, with APA citations and a downloadable BibTeX file.
Why you can trust the output
Statistics are computed by Python and R libraries. The AI recommends tests and smooths prose; it is never allowed to produce or restate a statistic, so it cannot hallucinate one.
Before export, KyroStat re-derives each reported p-value from its test statistic and degrees of freedom, the same consistency check journals apply with statcheck, and flags any mismatch.
Results include the effect size the test calls for, such as Cohen’s d, Hedges’ g, omega squared, or Cramer’s V, with confidence intervals where available, as APA 7 requires.
Want the detail on every capability? Explore the features or see how KyroStat compares to SPSS and GraphPad Prism.
Create an account and run your first analysis in minutes. No install, no statistics course required.