PIs and lab heads
One workspace for the lab’s analyses, with every result carrying the code that produced it.
Labs and departments
KyroStat for labs and departments
Your students and researchers run the right test, report it in APA 7 and keep the code. You review their work before it counts as done. Start with a free pilot for one cohort.
Who it is for
For the people who answer the statistics questions
Thesis supervisors
Students run the right test and write it up in APA 7; you review their work before it counts as done.
Departments and methods coordinators
A cohort of master’s or PhD students on one licence, with members approved by the department.
Writing centres and libraries
A tool to point students to when the statistics, not the prose, is what holds up the Results chapter.
Research networks
Members across institutions working on shared studies, with teams, deadlines and a shared paper.
What a lab gets
A study run inside KyroStat, not only the analysis
Everything an individual researcher gets, plus the parts a group needs and a single researcher does not.
Your organization, your members
Members join with your organization’s code and you approve each request. Approved members work without a personal paywall.
Teams with a lead
Group members into teams with one lead each and a size limit. Team leads are locked once a study starts.
Studies with a hard deadline
Each study has a deadline, tasks with due dates, dependencies between tasks, and fixed milestones from research question to submission.
Review before done
A member submits a task, a lead verifies it or requests changes with a note. Marking something complete is not the same as someone checking it.
A shared paper
The paper is split into sections, each owned by one member. A lead publishes versions, and every version is kept and can be restored.
Authorship report
For each project’s paper, a report shows which text KyroStat wrote and which the researcher wrote, with a snapshot kept at every export.
Versioned datasets
Members upload new versions of a study’s data. A lead decides which version is current, and earlier versions can be restored.
Reviewers and byline
Seat a statistician and a clinician as study reviewers, and keep an ordered author list with institutions and ORCID iDs.
Usage you can see
Admins see the organization’s AI usage and can set a monthly budget and per-member limits.
Each researcher gets
- A recommended test for the design, with the reason in plain words
- APA 7 tables and result sentences with effect sizes and confidence intervals
- The Python or R code that ran, with every result
- The p-value checked against the test statistic
- A PDF report with tables, plots, code and a Methods paragraph
How the numbers are checked
Same data, same model, same numbers
Given the same data and the same model, KyroStat gives the same numbers as Stata to the precision Stata prints. Where packages differ, it is because their defaults fit different models, not because one of them is wrong.
We do not write the test statistics
The arithmetic is done by SciPy, statsmodels and base R. KyroStat chooses the test, writes the code that calls them, and shows that code.
Checked against NIST certified values
On NIST’s ANOVA reference datasets, every one that 64-bit arithmetic can solve gives at least 7.4 correct significant digits, against the four decimals we report. Linear regression reproduces the certified coefficients, standard errors and R² to at least 6 digits on nine of NIST’s eleven datasets; a tenth reaches 5.77, and the eleventh is too ill-conditioned for any 64-bit package, a case KyroStat now flags.
Checked against published Stata output
On the hsb2 teaching dataset (n = 200), 35 statistics from an ANOVA, a t-test, two regressions and two chi-square tests agree with Stata’s published output to at least eight significant figures, and are identical at every precision a paper reports.
Where SPSS, Stata and R legitimately differ
These defaults only change the answer when groups are unequal in size or an assumption is violated. KyroStat writes the choice into the code it shows, so a reviewer can see which one was used.
| Setting | SPSS | Stata | R |
|---|---|---|---|
| Sums of squares (factorial ANOVA) | Type III | Type III | Type I |
| Independent t-test | Both rows | Pooled | Welch |
| Levene’s test centring | Mean | Median | Median |
| Chi-square continuity correction (2×2) | Yates | None | Yates |
What is not covered yet: the certification ran on the server’s Python stack. Analyses that run in the browser use newer versions of the same libraries, which have not been certified the same way yet. Effect sizes such as Hedges’ g and partial η² are computed by KyroStat and checked against their definitions rather than against another package. When a regression is too ill-conditioned for its digits to be trusted, the result carries a warning instead of passing silently.
Privacy
What happens to your group’s data
No rows go to the AI
The AI never receives the uploaded file or any individual row. To read the data and recommend tests it receives column names, types and summary statistics, and, for columns with up to 30 different answers, each answer and its count. Columns holding names or ID codes should be removed before upload.
Analyses run in the browser where they can
The code runs in the researcher’s browser on Python and R compiled to WebAssembly. When it cannot run there quickly, it runs on our servers in a single-use container with no network access.
Researchers control retention
The file is uploaded to our servers to be processed. A researcher can switch off server retention, which deletes the file once it is processed and the dataset once the project is completed. Projects can be deleted at any time.
What we do not claim
KyroStat holds no formal certification such as SOC 2 or ISO 27001. It is built for de-identified academic data, not identifiable clinical or regulated data.
The full detail is in the Privacy Policy and on the Security page.
Pricing
An annual licence, with a free pilot first
Start here
Free pilot
Free
One cohort, one thesis season
One supervisor’s students or one programme, with the full lab features, through one thesis season. No payment and no commitment to buy.
Annual licence
Quote on request
For a lab, a department or a network
One yearly price for your group, so approved members never meet a personal paywall. Tell us roughly how many people would use it and we will send a quote.
Book a 15-minute callBuying for yourself? See individual plans and the Thesis Pass.
Questions
We set up your organization and send you its join code. Members enter the code in their account settings and an admin in your group approves each request.
One cohort, such as one supervisor’s students or one master’s programme, through one thesis season, with the full lab features. There is nothing to pay and no commitment to buy afterwards.
Yes. Ask us and we will send the validation report and the test files, which re-run the comparison with one command.
It uses SciPy, statsmodels and R for the arithmetic, and it shows the code it ran. Students who already know SPSS can keep using it; KyroStat helps most with choosing the test, checking the reported numbers and writing them up.
Try it with one cohort first
Email support@kyrostat.com with your group and roughly how many students. We aim to reply within two business days.