One-way ANOVA, short for one-way analysis of variance, is a statistical method used to compare the mean values of three or more independent groups.
It is called “one-way” because there is only one grouping factor or independent variable. This factor divides the data into different groups. For example, a researcher may compare blood pressure across three treatment groups, exam scores across four teaching methods, or biomarker levels across several disease severity categories.
The main question answered by one-way ANOVA is:
Are all group means approximately the same, or is at least one group mean different from the others?
Unlike a t-test, which is designed for comparing two means, one-way ANOVA is used when the comparison involves three or more groups.
Why Is One-Way ANOVA Useful?
When researchers have more than two groups, it may seem natural to compare them using multiple t-tests. For example, if there are three groups, one could compare Group A with Group B, Group A with Group C, and Group B with Group C.
However, this approach creates a statistical problem. Each test has a chance of producing a false-positive result. If many tests are performed, the overall chance of finding a significant difference by chance alone increases.
One-way ANOVA avoids this problem by testing all group means in one overall analysis. Instead of asking whether each pair of groups differs, it first asks a broader question:
Is there evidence that any group mean differs from the others?
If the ANOVA result is statistically significant, researchers can then perform post-hoc comparisons to identify which specific groups are different.
How Does One-Way ANOVA Work?
Although ANOVA is often used to compare means, it works by analyzing variation in the data.
The total variation in the outcome is divided into two main parts:
1. Between-group variation
This represents how much the group means differ from the overall mean. If the group means are far apart, the between-group variation will be large.
2. Within-group variation
This represents how much individual values vary within each group. It reflects natural variability or random noise in the data.
One-way ANOVA compares these two sources of variation. It calculates an F-statistic, which is the ratio of between-group variation to within-group variation.
If the group means are very similar, the between-group variation will be small, and the F-statistic will be close to 1.
If at least one group mean is clearly different, the between-group variation will be larger than the within-group variation, leading to a larger F-statistic.
The test also provides a p-value. If the p-value is below the chosen significance level, commonly 0.05, the result is considered statistically significant.
What Does the Result Mean?
A statistically significant one-way ANOVA means that not all group means are equal. In other words, there is evidence that at least one group differs from the others.
However, ANOVA does not tell you exactly which groups are different. It only tells you that a difference exists somewhere among the groups.
To find out where the differences are, researchers usually perform post-hoc tests. Common post-hoc methods include Tukey’s HSD, Bonferroni correction, and other multiple-comparison procedures.
A non-significant ANOVA result means that the data do not provide enough evidence to conclude that the group means differ. It does not prove that all groups are exactly the same.
As with other statistical tests, the p-value should be interpreted together with effect size, confidence intervals, sample size, and practical or clinical importance.
Key Assumptions
Several assumptions should be considered before using one-way ANOVA.
1. Independent observations
The observations should be independent. Each participant, patient, or sample should belong to only one group. Measurements from one individual should not influence measurements from another.
If the same subjects are measured repeatedly across several time points or conditions, repeated measures ANOVA or a mixed-effects model may be more appropriate.
2. Continuous outcome variable
The dependent variable should be continuous, such as blood pressure, body weight, test score, laboratory value, reaction time, or symptom score.
3. Approximate normality within each group
The outcome should be approximately normally distributed within each group, especially when group sizes are small.
ANOVA is often reasonably robust to moderate departures from normality when sample sizes are balanced and not too small.
4. Homogeneity of variance
The variability of the outcome should be similar across groups. This assumption is often called homogeneity of variance.
If one group has much larger variability than another, the standard one-way ANOVA may be less reliable. Levene’s test is commonly used to assess equality of variances.
When variances are clearly unequal, Welch’s ANOVA may be a better choice.
5. No severe outliers
Extreme outliers can strongly affect group means and variances. It is good practice to inspect the data before running ANOVA.
When Should You Use It?
Use one-way ANOVA when:
- You have three or more independent groups.
- There is one grouping factor.
- The outcome variable is continuous.
- You want to compare the mean outcome across groups.
Common examples include:
- Comparing blood pressure across three treatment groups.
- Comparing test scores across different teaching methods.
- Comparing laboratory values across disease severity groups.
- Comparing anxiety scores across several intervention groups.
- Comparing weight loss across multiple diet programs.
When Should You Not Use It?
One-way ANOVA is not suitable when there are only two groups. In that case, an independent samples t-test is usually sufficient.
It is also not appropriate when the same participants are measured repeatedly across conditions or time points. For repeated measurements, repeated measures ANOVA or mixed-effects models should be considered.
If the outcome variable is categorical, such as improved/not improved or positive/negative, a chi-square test or logistic regression may be more suitable.
If the outcome is ordinal or highly non-normal, a non-parametric alternative such as the Kruskal-Wallis test may be considered.
What Happens After a Significant ANOVA?
A significant ANOVA result tells you that at least one group mean is different, but it does not identify the specific group differences.
Therefore, post-hoc tests are often needed after a significant result.
For example, if there are three groups: Group A, Group B, and Group C, post-hoc tests can compare:
- Group A vs Group B
- Group A vs Group C
- Group B vs Group C
These comparisons are adjusted to reduce the risk of false-positive findings caused by multiple testing.
Tukey’s HSD is commonly used when comparing all possible pairs of groups. Bonferroni correction is another widely used approach, especially when the number of planned comparisons is limited.
A Simple Example
Suppose a researcher wants to know whether different exercise programs lead to different weight loss outcomes.
Participants are randomly assigned to one of three groups:
- No structured exercise
- Moderate-intensity exercise
- High-intensity exercise
After 12 weeks, the researcher records weight loss for each participant.
The average weight loss is:
- No exercise group: 1.2 kg
- Moderate exercise group: 3.4 kg
- High-intensity exercise group: 5.1 kg
The researcher performs a one-way ANOVA and obtains:
- F(2, 57) = 8.42
- p = 0.001
Because the p-value is less than 0.05, the result is statistically significant. This suggests that average weight loss differs among the three groups.
The researcher then performs a post-hoc test and finds that the high-intensity exercise group lost significantly more weight than the no-exercise group. The moderate exercise group may or may not differ from the other groups depending on the post-hoc results.
How to Report the Result
A one-way ANOVA result can be reported like this:
“One-way ANOVA showed a statistically significant difference among groups, F(df_between, df_within) = value, p = value.”
For example:
“One-way ANOVA showed a significant difference in weight loss among the three exercise groups, F(2, 57) = 8.42, p = 0.001.”
A more complete report may include:
- Number of groups.
- Sample size in each group.
- Mean and standard deviation for each group.
- F-statistic.
- Degrees of freedom.
- p-value.
- Effect size, such as eta-squared or partial eta-squared.
- Post-hoc test results, if applicable.
Summary
One-way ANOVA is an important method for comparing the means of three or more independent groups. It provides an overall test of whether group differences exist while helping control the risk of false-positive findings that can occur when multiple t-tests are performed.
A significant ANOVA result indicates that at least one group mean differs from the others, but post-hoc tests are usually needed to identify the specific differences.
To use one-way ANOVA correctly, researchers should consider independence, normality, equal variances, outliers, and the need for post-hoc comparisons. The p-value should always be interpreted together with effect size and practical or clinical relevance.