One-Sample vs Two-Sample T-Test: What's the Difference?

The Short Answer: One Question Separates Them

Both tests belong to the same family of statistical procedures developed by William Sealy Gosset — who published as "Student" in 1908 — to test whether a sample mean differs from what you'd expect by chance alone. The difference between them is not mathematical complexity; it's what you're comparing. A one-sample test asks "does my group match this standard?" A two-sample test asks "do these two groups match each other?"

What Is a One-Sample T-Test?

The test was created to handle a practical problem: you have data from one group, and you want to know whether that group's true average equals some target, standard, or historical benchmark. The benchmark (μ₀) must come from outside your data — it could be a regulatory standard, a prior study result, a manufacturer's claimed specification, or a national average. You do not estimate it from the same sample.

Because you are estimating the population standard deviation (σ) from the sample standard deviation (s), the test statistic follows a t-distribution rather than the standard normal distribution. The shape of that distribution depends on degrees of freedom: df = n − 1. With larger samples, the t-distribution approaches the normal distribution, which is why the t-test and z-test give nearly identical results when n is large. See the full guide to the one-sample t-test for a deeper treatment.

When to Use the One-Sample T-Test

Use the one-sample t-test when all three conditions hold: you have a single group, the outcome is measured on a continuous scale, and you have a specific target value to compare against. Real-world situations that fit this pattern include:

• A teacher wants to know whether her class's average score (x̄ = 73) matches the district benchmark of 70.
• A manufacturer checks whether the mean fill volume of bottles (x̄ = 497 ml) matches the specification of 500 ml.
• A nutritionist tests whether the mean daily calorie intake of study participants equals the national recommendation of 2,000 kcal.
• A hospital checks whether its mean patient wait time matches the national average of 24 minutes.

One-Sample T-Test Formula

The numerator (x̄ − μ₀) measures the raw difference between your sample mean and the target. The denominator (s / √n) is the standard error — how much the sample mean would vary from sample to sample due to chance alone. Dividing by the standard error converts the raw difference into a t-score: how many standard errors away from μ₀ your sample mean sits.

Assumptions of the One-Sample T-Test

What Is a Two-Sample T-Test?

The two-sample t-test answers a different question than the one-sample version: not "does my group match a benchmark?" but "do these two groups differ from each other?" The key word is independent — the participants in Group 1 must have no relationship to the participants in Group 2. If the same people appear in both conditions (before/after, left eye/right eye), the correct test is the paired samples t-test, not the two-sample t-test.

There are two common versions of the two-sample t-test. Student's version assumes the two populations have equal variances and uses a pooled standard error. Welch's version drops that assumption and is recommended as the default in most modern statistical software — including R's default behavior — because it is robust when variances are unequal without losing much power when they are equal.

When to Use the Two-Sample T-Test

The two-sample t-test fits when you have two independent groups and want to know whether their means differ. Common applications include:

• Comparing exam scores between students taught with Method A and students taught with Method B (two separate classrooms).
• Testing whether a drug group and a placebo group have different average blood pressure readings.
• A/B testing: comparing average session time for users who saw Design A versus Design B.
• Comparing average salaries between male and female employees in the same role.
• Checking whether defect rates differ between two production lines.

Two-Sample T-Test Formulas

The Welch-Satterthwaite formula for degrees of freedom is:

df = (s₁²/n₁ + s₂²/n₂)² / [(s₁²/n₁)²/(n₁−1) + (s₂²/n₂)²/(n₂−1)]

This gives a non-integer df in most cases, which is handled automatically by statistical software. See the detailed breakdown of equal vs unequal variance t-tests for guidance on when each version applies.

Assumptions of the Two-Sample T-Test

Key Differences: One-Sample vs Two-Sample T-Test

Extended Comparison: How Related Tests Compare

Decision Guide: Which T-Test Should You Use?

The CHOOSE Framework for T-Test Selection

Worked Examples: Step-by-Step Calculations

Example 1 — One-Sample T-Test: Student Exam Scores

Example 2 — Two-Sample T-Test: Drug vs Placebo Blood Pressure

How to Run These Tests in Software

R

# One-sample t-test: compare sample mean to μ₀ = 70
scores <- c(78, 82, 65, 71, 80, 75, 88, 69, 74, 77)
t.test(scores, mu = 70, alternative = "two.sided", conf.level = 0.95)

# Output: t statistic, df, p-value, confidence interval, sample mean

# Welch's two-sample t-test (var.equal = FALSE is the default)
drug    <- c(15, 12, 18, 14, 13, 16)
placebo <- c(8, 10, 9, 7, 11, 6)
t.test(drug, placebo, alternative = "greater", var.equal = FALSE)

# For Student's (equal variances), set var.equal = TRUE

Python (SciPy)

from scipy import stats
import numpy as np

scores = np.array([78, 82, 65, 71, 80, 75, 88, 69, 74, 77])

# One-sample t-test against μ₀ = 70
t_stat, p_value = stats.ttest_1samp(scores, popmean=70)
print(f"t = {t_stat:.4f}, p = {p_value:.4f}")

from scipy import stats

drug    = [15, 12, 18, 14, 13, 16]
placebo = [8, 10, 9, 7, 11, 6]

# Welch's t-test (equal_var=False is the default)
t_stat, p_value = stats.ttest_ind(drug, placebo,
                                    equal_var=False,
                                    alternative='greater')
print(f"t = {t_stat:.4f}, p = {p_value:.4f}")

# Student's (equal variances): equal_var=True

Excel

Excel has built-in functions for both tests. Navigate to Data → Data Analysis → t-Test or use these formulas directly:

=== One-Sample T-Test ===
# Calculate t-statistic manually:
= (AVERAGE(A2:A26) - 70) / (STDEV(A2:A26) / SQRT(COUNT(A2:A26)))

# Two-tailed p-value:
= T.TEST(A2:A26, reference_array_equal_to_mu0, 2, 1)

=== Two-Sample T-Test ===
# Data Analysis → t-Test: Two-Sample Assuming Unequal Variances (Welch's)
= T.TEST(A2:A31, B2:B31, 1, 3)
# Tails: 1=one-tailed, 2=two-tailed
# Type: 1=paired, 2=equal variance, 3=unequal variance (Welch's)

For a guided walk-through, see the interactive t-test calculator on Statistics Fundamentals, which handles both test types with step-by-step output.

Interactive T-Test Calculator

Common Mistakes and How to Avoid Them

Effect Size: Cohen's d for Both Tests

A statistically significant result tells you the observed difference is unlikely to be due to chance. It says nothing about whether the difference matters in practice. Effect size addresses that gap. For t-tests, Cohen's d is the standard measure.

Cohen's d thresholds were proposed by Jacob Cohen in Statistical Power Analysis for the Behavioral Sciences (1988). They are widely cited and used across disciplines, including the APA's guidelines on reporting statistical results. For full worked calculations and interpretation, see the Cohen's d guide and the effect size calculator on Statistics Fundamentals.

How to Report Results (APA Format)

Both tests share the same reporting template. Report the test type, t-statistic, degrees of freedom, p-value, confidence interval, and effect size. The APA Style Guide for statistics sets the standard most journals and academic programs follow.

Real-World Applications by Field

Entity and Formula Glossary

Frequently Asked Questions