Paired Samples T-Test: When to Use, When Not to, and Step-by-Step Data Analysis in SPSS
The Paired Samples T-Test is a statistical test used to check for significant differences between the means of two measurements (dependent observations) from the same group. In this guide, we will explore when to use the Paired Samples T-Test, when to avoid it, and how to perform data analysis in SPSS with an example, in English.
When to Use Paired Samples T-Test?
The Paired Samples T-Test is used when:
- Comparing Two Related Measurements: You have two measurements from the same group, such as pre-test and post-test scores, or results from two different conditions.
Example: Is there a difference in test scores before and after training? - Dependent Variable is Continuous: The variable being measured (e.g., test scores, blood pressure, reaction time) must be continuous.
- Data is Paired: Observations must be paired, meaning two measurements from the same participant or unit (e.g., a student’s pre-test and post-test score).
- Differences are Normally Distributed: The distribution of paired differences (measurement 1 – measurement 2) should be approximately normal.
- No Extreme Outliers in Differences: There should be no extreme outliers in the paired differences.
Examples:
- Is there a significant difference in test scores before and after a training program?
- Is there a difference in blood pressure readings before and after taking a medication?
When Not to Use Paired Samples T-Test?
- Non-Normal Differences: If the paired differences are significantly non-normal (Shapiro-Wilk test Sig. < 0.05), the T-Test may not be reliable.
Alternative: Use a non-parametric test like the Wilcoxon Signed-Rank Test. - Categorical Dependent Variable: If the dependent variable is categorical (e.g., yes/no, pass/fail), the T-Test is not applicable. Use McNemar’s Test instead.
- Presence of Outliers: If there are extreme outliers in the paired differences, the T-Test results may be biased.
- Independent Data: If the measurements are independent (e.g., scores from two different groups), use the Independent Samples T-Test.
- Three or More Measurements: If you want to compare three or more related measurements, use Repeated Measures ANOVA.
- Missing Data: If there is significant missing data in paired observations, the T-Test results may be unreliable.
Example
Suppose we are conducting a study to check whether there is a significant difference in math test scores before (pre-test) and after (post-test) a training program.
- Dependent Variables: Pre-Test Score and Post-Test Score (continuous).
- Participants: 10 students from the same group.
Data:
Student | Pre-Test | Post-Test |
---|---|---|
1 | 75 | 80 |
2 | 78 | 82 |
3 | 80 | 85 |
4 | 82 | 88 |
5 | 79 | 83 |
6 | 77 | 81 |
7 | 81 | 86 |
8 | 76 | 79 |
9 | 83 | 87 |
10 | 80 | 84 |
We will now perform the Paired Samples T-Test in SPSS to check if there is significant improvement in scores after the training program.
Step-by-Step Process in SPSS
Step 1: Data Entry in SPSS
- Open SPSS: Launch SPSS software and create a new dataset.
- Define Variables:
- Variable 1: Pre_Test (Type: Numeric, Label: Pre-Test Score).
- Variable 2: Post_Test (Type: Numeric, Label: Post-Test Score).
- Enter Data:
- Enter the scores provided above into the Pre_Test and Post_Test columns.
The data view will look like this:
Step 2: Check Assumptions
It is essential to check the assumptions for the Paired Samples T-Test:
- Continuous Dependent Variables: Pre-Test and Post-Test scores are continuous, so this assumption is met.
- Paired Data: Each student’s pre-test and post-test scores are paired, so this assumption is met.
- No Significant Outliers in Differences:
- Create a boxplot for paired differences (Post_Test – Pre_Test):
- .
- Target Variable: Diff_Score.
- Numeric Expression: Post_Test – Pre_Test.
- Click OK.
- select.
- Drag Diff_Score to the Y-axis.
- Check the output. If there are no extreme outliers, the assumption is met.
- Create a boxplot for paired differences (Post_Test – Pre_Test):
- Normally Distributed Differences:
- Run the Shapiro-Wilk test for differences:
- .
- Add Diff_Score to the Dependent List.
- Click the Plots button, select Histogram and Normality plots with tests, and uncheck Stem-and-leaf.
- Click OK.
- Check the Tests of Normality table in the output:
Shapiro-Wilk Statistic df Sig. 0.962 10 0.786 Interpretation: Sig. = 0.786 (> 0.05), so the differences are approximately normal. The normality assumption is met.
- Run the Shapiro-Wilk test for differences:
Step 3: Perform Paired Samples T-Test
- Go to Menu:
- Click .
- Select Variables:
- The Paired-Samples T Test dialog box will open.
- Select Pre_Test and Post_Test and add them to Pair 1:
- Pre_Test to Variable 1.
- Post_Test to Variable 2.
- Use the right arrow to create the pair.
- Set Options:
- Click the Options button.
- Keep the Confidence Interval Percentage at 95% (default).
- Missing Values: Ensure “Exclude cases analysis by analysis” is selected.
- Click Continue.
- Click OK:
- After confirming all settings, click OK. The results will appear in the SPSS output window.
Step 4: Interpret Output
The SPSS output will show three main tables:
Paired Samples Statistics:
This table shows the mean, standard deviation, and sample size for both variables.
Pair 1 | N | Mean | Std. Deviation | Std. Error Mean | |
---|---|---|---|---|---|
Pre_Test | 10 | 79.10 | 2.726 | 0.862 | |
Post_Test | 10 | 83.50 | 3.028 | 0.957 |
Interpretation: The mean of Post-Test (83.50) is higher than Pre-Test (79.10).
Paired Samples Correlations:
This table shows the correlation between pre-test and post-test scores.
Pair 1 | N | Correlation | Sig. | |
---|---|---|---|---|
Pre_Test & Post_Test | 10 | 0.943 | 0.000 |
Interpretation: Correlation = 0.943 (strong positive correlation), Sig. = 0.000 (< 0.05), indicating that pre-test and post-test scores are highly correlated.
Paired Samples Test:
This table shows the T-Test results.
Pair 1 Pre_Test – Post_Test | ||||||
---|---|---|---|---|---|---|
Mean Difference | Std. Deviation | Std. Error Mean | t | df | Sig. (2-tailed) | |
-4.400 | 1.265 | 0.400 | -11.000 | 9 | 0.000 |
95% Confidence Interval of the Difference: Lower: -5.305 Upper: -3.495
Interpretation:
- Mean Difference = -4.400 (Post-Test scores are on average 4.4 points higher than Pre-Test).
- t-value = -11.000, df = 9, p-value (Sig. 2-tailed) = 0.000.
- p-value (0.000) < 0.05, so the null hypothesis (that pre-test and post-test means are equal) is rejected.
- 95% Confidence Interval: The true mean difference is likely between -5.305 and -3.495.
Conclusion: There is a statistically significant improvement in math test scores after the training program.
Step 5: Report Results (APA Style)
To report results in APA format, you can write:
A Paired Samples T-Test revealed that students’ math test scores after the training program (M = 83.50, SD = 3.03) were significantly higher than pre-test scores (M = 79.10, SD = 2.73), t(9) = -11.000, p < 0.001, d = 3.48.
Note: Cohen’s d (effect size) = |Mean Difference| / Std. Deviation of Differences = 4.4 / 1.265 ≈ 3.48, indicating a large effect size.
Step 6: If Assumptions Are Violated
If the normality assumption is violated (Shapiro-Wilk Sig. < 0.05 for differences), the Paired Samples T-Test may not be reliable. In this case, use the Wilcoxon Signed-Rank Test (non-parametric alternative).
Wilcoxon Signed-Rank Test in SPSS:
- Go to Menu:
- .
- Select Variables:
- Select Pre_Test and Post_Test.
- Pair 1: Add Pre_Test to Variable 1 and Post_Test to Variable 2.
- Select Test Type:
- Check Wilcoxon.
- Click OK.
- Interpret Output:
Test Statistics Post_Test – Pre_Test Z -2.803 Asymp. Sig. (2-tailed) 0.005 Interpretation: Sig. = 0.005 (< 0.05), the null hypothesis is rejected. Post-Test scores are significantly different from Pre-Test scores.
APA Style: “A Wilcoxon Signed-Rank Test indicated that math test scores after the training program were significantly higher than pre-test scores, Z = -2.803, p = 0.005.”
Tips
- Checking Assumptions is Crucial: Always check for normality and outliers in differences.
- Sample Size: For small samples (<30), the normality assumption is critical. For larger samples, the T-Test is more robust.
- Visualize Data: Create histograms or boxplots for paired differences ( ).
- Backup Data: Keep backups of data and output.
- Practice: Practice Paired Samples T-Test and Wilcoxon Signed-Rank Test with different datasets.