How to Run a t-Test on Mac & Windows — Without R or Python

A t-test answers one of the most common questions in data analysis: do two groups really differ, or is the gap just noise? Here's how to run one in Stratum in under a minute — and how to read what comes back.

Say you're looking at a clinical trial: a treatment group and a placebo group, and an outcome measured for each patient. The treatment group's average looks higher — but averages wobble. A two-sample t-test tells you whether that difference is large enough, relative to the spread in the data, to take seriously.

The one-minute version

Open your dataset in Stratum (drag the CSV onto the window).
Choose Analyze ▸ Two-Sample Comparison.
Pick your measurement column (the numeric outcome) and the grouping column that splits the rows into two groups.
Read the result: the difference in means, the t-statistic, the degrees of freedom, and the p-value.

Stratum's two-sample comparison report showing a t-test — Stratum's Two-Sample Comparison report — the t-test, in one click.

How to read the result

The p-value is the headline. It's the probability of seeing a difference at least this large if the two groups truly had the same mean. Small p-values argue against "no difference":

p < 0.05 — by the usual convention, the difference is "statistically significant." The groups likely differ.
p ≥ 0.05 — you don't have enough evidence to claim a difference. That's not proof they're the same; it may just mean a small sample.

But don't stop at the p-value. The difference in means (and its confidence interval) tells you how big the effect is — statistical significance and practical importance are not the same thing.

Equal variances, or not? Stratum reports Welch's t-test by default, which doesn't assume the two groups have the same variance. It's the safer choice in almost every real situation, and it's what modern statistical practice recommends.

Check the assumptions

A t-test assumes each group is roughly normally distributed. Stratum includes a normality check right in the report, and you can eyeball it with a Q-Q plot. If your data is badly skewed or your samples are small, switch to the Wilcoxon rank-sum (Mann–Whitney) test — a nonparametric alternative that's one click away in the same panel.

Paired data? If each row pairs a before-and-after on the same subject, use a paired test instead — it's a separate option in the Two-Sample panel and it's far more powerful for matched measurements.

Go deeper

This post is the express lane. For the full walkthrough — Welch vs. pooled, paired tests, the nonparametric route, and how to export the result — see lesson 14 of the tutorial.

Download Stratum Read the full lesson →