A worked example
One study, from question to chapter.
A real-shaped MSc study, followed the whole way. Every panel below is what ZStat actually produces at that step, not a picture of it.
Research question
Does digital literacy predict academic performance among postgraduate students at a Nigerian university?
Cross-sectional survey · target 150 respondents · one supervisor · six weeks
Build the questionnaire
Four items measuring digital literacy on a five-point scale, plus CGPA and study hours. ZStat reads the wording as you write it.
I can evaluate whether an online source is credible
"Do you agree that digital skills are essential for academic success?" tells the respondent which answer is wanted. The instrument cannot be locked until it changes.
Try to open collection
It refuses, and says exactly what is missing.
- No ethics approval reference recorded.
- The stated purpose is four words. Consent must be specific enough for a respondent to know what they are agreeing to.
- No retention date. "Until the research is complete" is not a schedule anyone can act on.
Fill those in and the gate opens. This is the step most software leaves to a policy document nobody reads again.
Share one link
It opens on a phone, shows the consent notice before any question, and gives each respondent a code they keep.
Check the scale before trusting it
The digital literacy scale was reliable, Cronbach's alpha = .84, 95% CI [.79, .88], four items, N = 150.
No item raises alpha if removed. Item-total correlations .58 to .71.
Ask the question, and get told when you asked wrongly
A t test compares two groups. "Faculty" has four: Science, Arts, Management, Engineering. Use a one-way ANOVA.
Digital literacy significantly predicted CGPA, B = 0.31, SE = 0.07, t(148) = 4.43, p < .001, 95% CI [0.17, 0.45], accounting for 11.7% of the variance (adjusted R² = .111).
| Predictor | B | SE | beta | t | p |
|---|---|---|---|---|---|
| (Constant) | 2.41 | 0.24 | 10.04 | <.001 | |
| Digital literacy | 0.31 | 0.07 | .34 | 4.43 | <.001 |
| Study hours | 0.02 | 0.01 | .14 | 1.78 | .077 |
The residuals are mildly right-skewed (Shapiro-Wilk p = .03). With N = 150 the estimates are robust to this, but report it. The Durbin-Watson statistic of 1.94 shows no serial correlation, and no predictor has a VIF above 1.3.
Export the chapter
A Word document with the tables formatted, the sentences written, and a Method section assembled from what the study already recorded — the design, the ethics reference, the locked instrument, the participant flow.
Participants were 150 postgraduate students recruited between 3 March and 9 March 2026. Of 163 approached, 150 completed the questionnaire and 13 were excluded for incomplete responses. The Digital Literacy Questionnaire version 1.0 was administered, comprising four items across one construct on a five-point Likert scale.
Send it to your supervisor
A read-only link. No account, no password, expires on a date you set, withdrawable.
Every result computed by ZStat and recorded when it was run, with the SHA-256 digest of the data it consumed. Nothing was typed by hand, and no language model produced any number, interpretation or citation.
Six months later, a respondent withdraws
They enter the code they were given. Their answers are deleted; a record that a withdrawal happened is kept, with nothing that identifies them.
1 respondent has withdrawn and their data has been erased (4 answers removed). 3 analysis records predate the withdrawal. They were correctly computed on the data that existed when they were run, but their digests will no longer reproduce and their n no longer matches this study. Re-run them before anything is written up from them.
This is the part that turns a thesis defence from an argument into a demonstration.
Your study, in the same shape.
Free for the first one, with no time limit and no card.