Repeated Measures ANOVA in SPSS: How to Run, Read, and Report It

You open your SPSS file and see one group of participants measured before, during, and after an intervention, but you are not sure which test to use. You need to distinguish this situation from one-way ANOVA in SPSS, which is usually used when the groups are independent. Repeated Measures ANOVA is appropriate when the same participant appears at several measurement times or under several conditions.

The workflow below follows the order you need: identify the test, check the assumptions, set up the data, run the analysis in SPSS, read the output tables, and write the result. The numerical tables are labelled illustrative output. They are included to help you recognize the output, not to present findings from a real study.

What Repeated Measures ANOVA tests

Repeated Measures ANOVA in SPSS tests whether the mean of a quantitative variable differs across three or more measurement occasions or conditions when those measurements come from the same participant. For example, you may measure the satisfaction level of 30 students at three time points: before the course, after the course, and one month later.

Each row usually represents one respondent, while the columns Diem_Truoc, Diem_Sau, and Diem_TheoDoi contain the three measurements for that person. Because the measurements are dependent, you should not treat them as three independent groups. If you have only two measurements, the Paired-Samples T Test is often the more direct option. You can also review independent samples t test in SPSS when your design contains two independent groups.

Repeated Measures ANOVA answers a question about the overall mean difference. If the overall test is significant, you need pairwise comparisons to determine which time points differ. A small p-value does not tell you by itself which groups are different.

The design can also be extended when you have one repeated-measures factor and one group factor, such as scores measured at three time points for an intervention group and a control group. The model can then test the time effect, the group effect, and the interaction between time and group. This is often called mixed ANOVA. It differs from one-way ANOVA in SPSS, which has only one grouping factor.

H0 and H1 hypotheses

H0 states that the dependent variable has the same mean across the measurement occasions or conditions being compared. With three time points, H0 can be expressed as the population means for the pre-test, post-test, and follow-up scores being equal.

H1 states that at least one mean differs from another. H1 does not identify the specific time points that differ. You can draw a conclusion about a particular pair only after examining the pairwise comparison test.

Set the significance level before reading the results. It is often 0.05 when your research proposal specifies this level. The interpretation rule is as follows: if the p-value is smaller than the significance level, reject H0. If the p-value is greater than or equal to the significance level, there is insufficient evidence to reject H0. Do not write that H0 is true when the p-value is large, because the test only shows that the available evidence is not strong enough to reject H0.

Assumptions and conditions

Before looking up how to run Repeated Measures ANOVA in SPSS, check the design and the data. The dependent variable should be quantitative. The measurements must belong to the same participants and must be correctly paired using the respondent ID. If a participant has a score at the first time point but is missing a score at a later time point, SPSS may exclude that participant depending on the missing-data procedure being used.

The table below summarizes the main conditions and how to check them. For distributional assumptions, do not rely on a single number. Combine graphs, sample size, and the context of the data.

Sphericity is a specific assumption of Repeated Measures ANOVA. It concerns the variances of the differences between each pair of measurement occasions. With only two measurement levels, this assumption does not create the same issue as it does with three or more levels. Read Mauchly's Test of Sphericity before selecting the results row.

How to run Repeated Measures ANOVA in SPSS

You can run Repeated Measures ANOVA in SPSS using the steps below. Menu names may vary slightly across SPSS versions, but the main dialog remains Analyze > General Linear Model > Repeated Measures.

Step 1: Check and organize the data file

In the common SPSS data format, each respondent appears on one row. Create numeric variables such as Diem_T1, Diem_T2, and Diem_T3, with each variable representing one measurement occasion. Set the variable type to Numeric, check the missing-value codes, and make sure that high and low scores have the same meaning at every time point.

Do not place the time-point label and the score in one column if you want to run the analysis directly in the Repeated Measures dialog. Open Variable View to check variable labels, decimal places, and missing codes. If a variable is calculated from several items, finish calculating the score before running the test.

Step 2: Define the repeated-measures factor

Select Analyze > General Linear Model > Repeated Measures. In Within-Subject Factor Name, enter the factor name, such as ThoiDiem. In Number of Levels, enter the number of measurements, such as 3. Click Add, then select Define.

In the next window, move Diem_T1, Diem_T2, and Diem_T3 into the cells in the correct time-point order. The order matters because SPSS uses it to create the comparisons and labels in the output. If you have a grouping variable such as Nhom, move it to Between-Subjects Factor(s).

Step 3: Select descriptive tables and comparison tests

Click Plots, move ThoiDiem to the horizontal axis, select Add, and then select Continue. Click Options, then select Descriptive statistics, Estimates of effect size, and Compare main effects if your SPSS version displays this option. You can select the Bonferroni adjustment for pairwise comparisons.

If the model includes a grouping variable, also select Descriptive statistics and the appropriate estimates for between-group effects. Click Continue, then OK. When you need to save residuals or inspect graphs, use the Save button and select only the variables you actually need so that the file does not become cluttered.

Step 4: Save the syntax and record your choices

After running the analysis, open the Syntax window or select Paste instead of clicking only OK. Save the syntax with the data file. This lets you reproduce the result after changing a variable or explain to your supervisor why you used the Greenhouse-Geisser adjustment.

If your study has only one factor and the groups are independent, see how to run one-way ANOVA in SPSS to distinguish it from a repeated-measures design. If you have two factors, such as time and group, read more about the logic of ANCOVA in SPSS, but do not add a covariate unless your research proposal provides a theoretical basis for that variable.

Reading the output and drawing a conclusion

The output usually begins with Descriptive Statistics, followed by Mauchly's Test of Sphericity, Tests of Within-Subjects Effects, Pairwise Comparisons, and, when a grouping variable has been added, Tests of Between-Subjects Effects. Do not jump directly to the p-value without checking the corrected row.

The table below is illustrative output, presented in the format of SPSS output. The numbers only help you identify the columns and the logic of the conclusion. They are not findings from a real study.

In Mauchly's Test of Sphericity, if the Sig. column is smaller than 0.05, do not use the Sphericity Assumed row for your conclusion. Move to the Greenhouse-Geisser row, which SPSS usually displays in Tests of Within-Subjects Effects. If Mauchly's test is not statistically significant, you can read the Sphericity Assumed row, but still report clearly which table and row you used.

In Tests of Within-Subjects Effects, check the Sig. column for the ThoiDiem effect. In the illustrative output, p = 0.004, which is smaller than 0.05. You would therefore reject H0 and conclude that the mean scores differ across time points at the overall level. Then read Pairwise Comparisons to identify the differing pairs. When the table contains several rows, prioritize the Sig. column adjusted with Bonferroni or the method you selected.

If you have a grouping variable, also check the interaction effect in Tests of Within-Subjects Effects. It usually has a name combining the time factor and the grouping variable. A significant interaction indicates that the amount of change over time is not the same across groups. Reporting only the main effects can hide this important part of the model.

Repeated Measures ANOVA differs from how to run two-way ANOVA in SPSS because one factor can be repeated measures while the other is an independent group factor. State clearly in the research methods section which factor is the Within-Subjects Factor and which is the Between-Subjects Factor.

How to write this in your thesis

In Chapter 4, write the result in four parts: the means and standard deviations, the sphericity check, the overall effect test, and the pairwise comparisons. State which corrected row you used. Avoid writing that ANOVA proves an intervention caused a change if your design only measures participants before and after the intervention without a control group.

You can adapt the following template to your actual output: “Mauchly's test showed that the sphericity assumption [was met/was not met], p = [value]. Therefore, the Repeated Measures ANOVA result was read from the [Sphericity Assumed/Greenhouse-Geisser] row. The effect of [repeated-measures factor] was [statistically significant/not statistically significant], F([df1], [df2]) = [F], p = [p-value], [partial eta squared if reported]. Pairwise comparisons with the [Bonferroni] adjustment showed that [time-point pair] differed, p = [p-value].”

When the result is statistically significant, add the mean and standard deviation for each time point. If the software reports Partial Eta Squared, you can report it as an effect-size measure, but do not label it small, medium, or large unless your proposal specifies how it should be interpreted. The conclusion should return to the research question and the specific variable, rather than repeating only that “there was a difference.”

What to use when an assumption is violated

A violation of sphericity does not mean that you should delete the data or rerun the analysis until the p-value looks favorable. First, keep the model, read the Greenhouse-Geisser or Huynh-Feldt row, and state the choice in your thesis. The row you select should follow the test table and the analysis rule established in advance.

If the data are strongly skewed, contain serious outliers, or do not fit the assumptions of ANOVA, consider a nonparametric test for several repeated measurements, such as the Friedman Test. If the groups are independent and there are only two groups, Mann-Whitney in SPSS may be more suitable for some designs. With three or more independent groups, see Kruskal-Wallis in SPSS. These are different options for different designs, not a universal repair button for every data file.

If the problem concerns missing data, check how many participants remain at every measurement occasion and why values are missing. If missingness changes the sample size substantially between time points, report the limitation and discuss it with your supervisor. Do not fill cells with the mean just to complete the dataset unless an approved rule allows it.

Common mistakes

Choosing the wrong type of ANOVA

A common mistake is using one-way ANOVA for repeated scores from the same participant, or using Repeated Measures ANOVA when each time point contains a different group of participants. Draw a simple diagram: who is each row, what is each column, and in how many columns does one participant appear.

Reading the Sphericity Assumed row when Mauchly's p-value is small

This error changes the df and p-value. If Mauchly's Test of Sphericity has a Sig. value smaller than 0.05, move to the corrected row and state the reason in your report.

Concluding which groups differ from the ANOVA table alone

The overall test only says that a difference exists somewhere among the levels. You need to read Pairwise Comparisons, check the adjusted p-value, and report the mean for each time point.

Changing the variable order or using unclear names

If Diem_T1, Diem_T2, and Diem_T3 are not in the correct order, the Pairwise Comparisons table will be difficult to interpret. Name variables by time point, save the syntax, and check Descriptive Statistics before writing.

You can also review Levene's test in SPSS when the model includes a group factor and you need to examine variance between groups. Levene's test does not replace the sphericity check because the two tests examine different conditions.

Frequently asked questions

How many measurements can Repeated Measures ANOVA in SPSS handle

The test is commonly used with three or more measurements or conditions for the same participant. With two measurements, the Paired-Samples T Test is often easier to interpret, but the final choice must still follow the research design and proposal.

Do you need to check Levene's test for Repeated Measures ANOVA

Levene's test mainly concerns equal variances between independent groups. For one-factor Repeated Measures ANOVA without independent groups, focus on sphericity and the dependence structure between measurements. If the model includes a group factor, check the assumptions relevant to the between-groups component.

Which table should you read when Mauchly's Test has Sig. equal to 0.000

SPSS usually displays 0.000 when the p-value is extremely small. It does not mean that p is exactly 0. When Sig. is smaller than 0.05, the sphericity assumption is violated, so read the Greenhouse-Geisser row or another appropriate correction in Tests of Within-Subjects Effects.

What if the ANOVA p-value is significant but none of the pairwise comparisons is significant

This can happen because the overall test and the multiple-comparison adjustment have different sensitivity, or because the effect is distributed across the levels without a clear pairwise difference. Report both results. Do not omit Pairwise Comparisons or claim that a specific pair differs when its p-value does not reach the significance level.

Can you use Repeated Measures ANOVA with Likert data

State clearly how you created the dependent variable. If you use a total or mean score from several items, the test choice must fit the scale construction, sample size, and data assumptions. Do not look at one individual Likert item and assume that every ANOVA condition has been met.

Open the .sav file, check the order of the measurement columns, rerun the analysis using the saved syntax, and record three details separately: the sphericity row used, the p-value for the overall effect, and the result for each pair. If you want to run the analysis on your own data in SPSS or SmartPLS, you can use M4 data analysis by DoThesis to complete this step with your actual file.