Questions

Yes, and with no time limit. The free plan runs all 86 statistical procedures, every assumption check and every refusal, on one study of up to 150 responses. No card is needed. Paying buys more studies, larger samples, collaborators and an export without a footer line - never a different or better analysis.

You need to know what you are trying to find out. The catalogue is written in those terms - "compare three groups", "predict a yes or no" - rather than in procedure names. ZStat picks the test and tells you when the one you asked for is wrong. It will not tell you whether your research question is a good one.

Yes. CSV, Excel and SPSS .sav files. You can also collect fresh responses through ZStat, which is where the consent and withdrawal machinery applies.

Upload or collect, choose what you want to know, and the sentence comes back formatted. The slow part of a thesis is deciding what to ask, and no software fixes that.

It stops and says which one fits. A t test asked for on three groups names the three groups it found and points at the one-way ANOVA. This is the main thing ZStat does that a general statistics package does not.

Every time, whether or not you ask, and the failures are reported with the result rather than buried. Normality, equal variances, independence, linearity, multicollinearity, and the ones specific to each procedure.

Each procedure was gated against a reference implementation before it shipped, and the development log records what each was checked against. Every result also carries the SHA-256 of the data that produced it, so you can re-run it later and see whether anything changed.

No. No language model produces any number, interpretation or citation. That is stated in the assurance document because ethics committees ask.

It says so rather than running something approximate. Some specialised procedures are not here.

On the hosted service, on ZStat infrastructure, with each institution in its own separate database. Self-hosted, entirely on your own server - we hold no copy and could not produce it if asked.

You do. On the hosted service you are the data controller and ZStat is the processor, acting on your instructions.

Yes, using the code they were given at submission, with no account and no reason required. Their answers are deleted. A record that a withdrawal happened is kept - counts and dates, nothing that identifies them - so you can explain why your totals changed.

A scheduled job deletes what is past it, counting from when each respondent answered rather than from when the study was created. A sweep that would remove more than half a study stops and asks first.

In transit, yes. At rest, respondent codes are pseudonymous but stored in plain text, and the assurance document says so plainly rather than omitting it. Where a study declares sensitive categories, the compliance check blocks collection for that reason.

How much, never how correctly. Studies, responses per study, collaborators, and whether the exported document carries a footer line. Every check, refusal and assumption test is on all plans.

Charging for the part that stops you publishing something wrong would mean the people who cannot pay get a tool that runs an invalid analysis silently. That is the wrong way round.

Card through Paystack for individual plans. Departments and institutions are invoiced annually.

ZStat becomes read-only. You keep signing in, reading every result, exporting your data and reports, honouring withdrawals and running retention. What pauses is new work. Your data is never withheld - you remain the data controller and are legally obliged to be able to act on it.

Yes - your own subdomain, or point your own domain at us. Your institution gets its own database, separate from every other institution, so isolation is physical rather than a filter in a query.

Yes. A department becomes a unit inside your institution account, so when the university buys across the board later, nothing has to be migrated.

Yes. ZStat Self-Hosted is licensed annually through ztech.ng and runs inside your network. Respondent data never reaches us.

They create an account, then ask to join. An administrator at your institution approves. Saying you are at a university is not evidence that you are.

The assurance document, generated from the running installation and downloadable as Word. It reports what the software does, including a section on what is not the case.

There is a desktop application for Windows on the paid plans, for when the connection is unreliable or the data may not leave the room.

The collection page is built for phones first, which is where respondents answer. The analysis interface works on a phone and is more comfortable on a laptop.

You can issue a read-only link they open without an account. It shows the results and the assumption checks, expires on a date you set, and can be withdrawn.

Yes, and the codebook goes with it, so the variable names still mean something in the other tool.

Tell us. The development log records the defects found so far and what each one broke, because software that hides its bugs is not the kind you should trust with a thesis.