METHOD GUIDE
Independent Samples t-Test in SPSS
Which Row to Report and Why Welch Is Safer
SPSS prints two rows of results for this test, and the row you report changes the p-value and the interval. This guide covers the design checks, how to read the output, and how to report the mean difference in APA 7 with an effect size and its interval.
IN SHORT
What the Test Estimates and Which Row Answers It
An independent samples t-test compares the mean of a continuous outcome between two unrelated groups. It estimates the mean difference with a confidence interval and tests whether that difference is compatible with zero. It fits one numeric outcome and a grouping variable with two categories, where each participant belongs to one group only. To run an independent samples t-test in SPSS, open Analyze, Compare Means (Compare Means and Proportions in recent versions), Independent-Samples T Test, define the two group codes, and report the row labeled Equal variances not assumed unless pooling was justified from the design in advance. OnlineSPSS, a specialist statistical consultancy for academic, clinical and professional research, carries out and reviews this analysis through an in-house PhD-level statistician.
ANALYSIS SUPPORT
Analysis Support for This Method
Reading the two rows is one part of the job. For the comparison carried out and reported on your own data, or for an outcome better compared by ranks, see the two pages below.
Independent Samples t-Test
The comparison of two independent group means carried out within your study design, with the diagnostics, the effect size and its interval, and the reporting agreed for your data.
Mann-Whitney U Test
A rank-based comparison of two independent groups, for an ordinal outcome or when the mean is not the right summary of the difference between them.
BEFORE YOU RUN IT
Four Decisions Before the First Click
Four decisions made before the first click decide whether the numbers SPSS prints answer your question, and none of them can be repaired afterwards by reading the output more carefully.
Two Separate Groups Each participant belongs to one group and is measured once. Siblings, classes or clinics create dependence the test cannot see, and call for a paired or multilevel approach.
A Continuous Outcome The outcome should be measured on an interval or ratio scale, or a rating scale treated as approximately continuous with that choice stated. A clearly ordinal outcome may suit a rank-based test.
Shape and Outliers Plot the outcome in each group instead of relying on a single test. Moderate or large groups tolerate mild skew, while strong skew or influential outliers in small groups need a reasoned response.
Spread in Each Group Levene's test in SPSS is printed beside the t-test (from version 31 only on request), but it should not choose the method. Screening first and then picking a row distorts error rates, so the unequal-variance row is the default.
IN SPSS
How to Run the Independent Samples t-Test in SPSS
The outline below is for IBM SPSS, and menu labels differ slightly between versions, so the decisions around the steps matter more than the clicks.
- Confirm the design: one continuous outcome, two independent groups, each case counted once.
- Plot the outcome by group and look at the spread and any extreme values before testing.
- Open Analyze, Compare Means (Compare Means and Proportions in recent versions), Independent-Samples T Test, and place the outcome under Test Variable(s) and the group variable under Grouping Variable.
- Use Define Groups to enter the two codes, and check them against the value labels.
- Keep the 95% confidence interval, and in recent versions request the effect size estimates.
- Run it and read the group table first, then the test table, then the row you decided to report.
READING THE OUTPUT
What Each Part of the SPSS t-Test Output Tells You
One table holds two rows of results, and the check that matters most is that the t, the degrees of freedom and the interval all come from the same row.
| Output element | What it tells you | What to check |
|---|---|---|
| Group Statistics | The N, mean, standard deviation and standard error of each group | That each N matches the intended sample and the labels are the right way round |
| Levene’s Test for Equality of Variances (on request from version 31) | Whether the spread of the outcome differs between the groups | That it is treated as a signal about the data, not as the switch between rows |
| Equal variances assumed row | The t statistic, its degrees of freedom and p from a pooled variance | That it is chosen on design grounds, such as equal group sizes, and not on Levene’s p |
| Equal variances not assumed row | The same quantities with Welch’s adjusted degrees of freedom, usually not whole numbers | That the degrees of freedom are reported as printed, to one or two decimals |
| Mean Difference and Std. Error Difference | The estimated difference between the groups and its precision | That the sign follows the Group 1 minus Group 2 order set in Define Groups and matches your sentence |
| 95% Confidence Interval of the Difference | The values for the difference that the data are compatible with | That it comes from the same row as the t statistic and is stated in outcome units |
| Independent Samples Effect Sizes (recent versions) | Cohen’s d and Hedges’ correction, each with an interval | That the standardizer is named, since the pooled one sits beside a Welch test |
REPORTING
The APA 7 Template for a Two-Group Comparison
Treat the sentence below as a frame: every blank is filled from your own output, and nothing in it is a result.
Write the core as t(df) = ___, p = ___, followed by the mean difference with its 95% CI [___, ___] and a standardized effect such as Hedges’ g with its interval. Give the mean and standard deviation of each group and the group sizes, and say in words which group scored higher and by how much in the units of the outcome. Report the degrees of freedom of the row you used and state that equal variances were not assumed, which for Welch’s version usually gives a non-integer value. Name the coding of the groups so that the sign of the difference is clear, and keep the SPSS group statistics table as the source of the descriptives.
COMMON MISTAKES
Mistakes Reviewers Find in Reported t-Tests
Reviewers return to the same six points when they query a two-group comparison, and most trace back to the two rows of the test table.
Row Chosen After Seeing p Picking the row that gives the smaller p-value is an undeclared analytic choice. The rule for the row is set from the design before the output is read, and it is stated in the text.
The Wrong Sign SPSS subtracts Group 2 from Group 1 as entered under Define Groups, so a negative difference can describe the group you call higher. Check the sign against the two group means.
Pooled Degrees of Freedom Degrees of freedom equal to the total sample size minus 2 belong to the pooled row. Printing them beside the Welch statistic mixes two rows of the table, and reviewers notice.
Significance as Size A p-value does not say how large a difference is. The difference in outcome units with its interval, and a standardized effect, let the reader judge whether it matters in practice.
Groups Not Independent Two groups drawn from the same classes, wards or families do not provide independent observations, so the standard errors come out too small. The design tells you this, not a diagnostic.
Many Outcomes, One Threshold A dozen outcomes tested at .05 will yield a few small p-values by chance. Outcomes are prespecified, or the multiplicity is handled and the number of tests is stated.
Already run this analysis? Independent Statistical Review checks the row you reported, the test choice and the write-up before you submit.
FREQUENTLY ASKED QUESTIONS
Questions About the Independent Samples Output
Short answers to what researchers ask when two rows appear on screen, from the choice of row and the effect size to confidentiality and the assessment.
Should I report the equal variances assumed or not assumed row?
Report the equal variances not assumed row by default, which is Welch’s version of the test. It performs almost as well as the pooled version when variances are equal and protects the error rate when they are not, especially with unequal group sizes. The pooled row can be justified in advance by a design with equal group sizes and similar spread, but a Levene p-value above .05 is not a good reason to switch.
Why are the degrees of freedom in my SPSS output not a whole number?
The Welch row adjusts the degrees of freedom for unequal variances with the Welch-Satterthwaite approximation, and the result is rarely an integer. Report the value as printed, to one or two decimals. Do not round it, and do not replace it with the pooled value of the total sample minus 2. The adjusted value is also the one that matches the p-value and interval in the same row.
Which effect size should I report with the mean difference?
Report the mean difference in outcome units with its interval first, because it keeps the measurement scale. A standardized effect, Hedges’ g or Cohen’s d with its interval, adds comparability across studies. Recent SPSS versions print both in a separate table, and the text should say which standardizer was used.
Can a statistician check which row I reported before I submit?
Yes. An in-house PhD-level statistician can review the output you already have and confirm that the reported row, degrees of freedom and effect size agree, or carry out the comparison on your data from the start. If the analysis is settled and only the tables need work, Results Reporting & Interpretation covers that stage. The interpretation and the conclusions remain yours.
How do you protect my research data and confidentiality?
Your data and project details are handled confidentially, and project information is not published. Researcher names and institutions are not used publicly without permission. An NDA is available where required and can be arranged before sensitive materials are shared. Read more about confidentiality and data security.
What happens when I request an assessment, and how long does it take?
You share your research question, design and relevant materials, such as the dataset or the output you have, and a statistically trained Research Coordinator reviews the request and responds the same day. The analysis, deliverables, timeline and one fixed project fee are then confirmed before any work begins, so the timeline follows from the agreed scope.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
Need Support With Your
Group Comparison?
Share your research question, how the outcome and the groups were measured and the output you have so far. We confirm the test and the reporting for your design before any work begins.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014