Two-Sample T-Test (Independent Samples)
Compare the means of two independent groups to see if they differ significantly. Choose Student's t-test for equal variances or Welch's t-test for unequal variances, with the t-statistic, p-value and confidence interval.
Run Two-Sample T-Test →What is a Two-Sample T-Test?
A two-sample t-test, also called an independent-samples t-test, compares the means of two separate, unrelated groups to determine whether they differ significantly. Examples include comparing output from two machines, results from two suppliers, or a treatment group against a control group.
There are two versions. Student's t-test assumes the two groups have equal variances and pools them into a single estimate. Welch's t-test does not assume equal variances and is more robust when they differ; because unequal variances are common, many practitioners use Welch's test by default.
The test computes a t-statistic from the difference between the two group means relative to the variability within them, and converts it to a p-value. A small p-value indicates the two group means differ by more than sampling variation would explain.
In plain terms: You have two separate groups and want to know if their averages really differ. The test weighs the gap between the averages against how noisy each group is. If the groups have very different spreads, use the Welch version, which handles that safely.
Two Versions
Student's t-test
Assumes the two groups have equal variances and pools them. Slightly more powerful when the equal-variance assumption truly holds.
Welch's t-test
Does not assume equal variances and adjusts the degrees of freedom accordingly. A safe default, since unequal variances are common.
Independent Groups
The two samples must be independent. For paired or matched data, use the paired t-test instead.
Key Formulas
Choosing and Interpreting
If Levene's or an F-test suggests the variances are unequal, or if you are unsure, use Welch's t-test; it costs little power when variances are equal and protects you when they are not. Compare the p-value to your significance level to decide on the difference.
The confidence interval for the difference in means is the most useful output: if it excludes zero, the difference is significant, and its width and location show how large and in which direction the difference is.
Assumptions & Validation
Independent Samples
The two groups are independent of each other.
If violated: For paired or repeated measures, use the paired t-test.
Approximate Normality
Each group is approximately normal, or samples are large enough for the Central Limit Theorem.
If violated: Use the Mann-Whitney U test for small, non-normal samples.
Variance Assumption
Student's version assumes equal variances; Welch's does not.
If violated: Use Welch's t-test when variances differ or are unknown.
⚠️ Check assumptions first
The most common pitfall is applying the equal-variance Student's t-test when the two groups actually have unequal variances, which distorts the p-value. When variances are unequal or unknown, Welch's t-test is the safer choice and loses almost no power when variances happen to be equal. Also ensure the samples are truly independent; paired data requires the paired t-test, not this one.
When NOT to Use Two-Sample T-Test
Paired Data
For before-and-after or matched pairs, use the paired t-test, which is more powerful for related measurements.
Three or More Groups
To compare more than two group means, use ANOVA to control the overall error rate.
Non-normal Small Samples
For small, clearly non-normal groups, use the Mann-Whitney U test.
Industry Applications
A/B Comparison
Compare a metric between two variants, machines, or suppliers to see which performs differently.
Treatment vs Control
Test whether a treatment group's mean outcome differs from a control group's.
Process Comparison
Compare a quality characteristic across two lines, shifts, or materials.
Before-After Across Groups
Compare independent groups measured under two different conditions.
Frequently Asked Questions
What is the difference between Student's and Welch's t-test?
Student's t-test assumes the two groups have equal variances and pools them into one estimate, while Welch's t-test does not assume equal variances and adjusts the degrees of freedom to account for the difference. Because unequal variances are common and Welch's test loses very little power when variances are actually equal, many statisticians recommend using Welch's t-test as the default.
When should I use a two-sample t-test instead of a paired t-test?
Use the two-sample t-test when the two groups are independent, meaning the observations in one group are unrelated to those in the other, such as different machines or separate treatment and control groups. Use the paired t-test when the measurements are related, such as before-and-after readings on the same subjects. Applying the wrong one either wastes power or invalidates the test.
How many groups can a two-sample t-test compare?
Exactly two. To compare the means of three or more groups, use analysis of variance (ANOVA), which tests all groups simultaneously while controlling the overall Type I error rate. Running multiple t-tests across many groups inflates the chance of a false positive, which is precisely the problem ANOVA is designed to avoid.
What if my data is not normally distributed?
The two-sample t-test is reasonably robust to mild non-normality, especially with larger samples, because of the Central Limit Theorem. For small samples that are clearly non-normal or contain outliers, the Mann-Whitney U test is a non-parametric alternative that compares the two groups using ranks rather than means and does not assume normality.
How do I interpret the confidence interval for the difference?
The confidence interval gives the plausible range for the true difference between the two group means. If the interval excludes zero, the difference is statistically significant at the corresponding level. The location and width of the interval also convey the size and direction of the difference, which matters for deciding whether a significant result is practically meaningful.
Should I test for equal variances first?
You can test equality of variances with an F-test or Levene's test, but a simpler and increasingly recommended approach is to use Welch's t-test by default, which does not require that assumption. This avoids the pitfalls of a preliminary variance test influencing the main analysis, and Welch's test performs almost as well as Student's when variances happen to be equal.
Compare Two Independent Group Means
Run Student's or Welch's two-sample t-test with full output. Free during Beta.
Run Two-Sample T-Test →