Normality Tests in SPSS

Summary

In this video, Dr. Grande explains how to test for normality in SPSS, using pre-test and post-test scores from a study with two groups: control and CBT. He demonstrates how to use the 'Explore' function in SPSS to generate histograms and normality plots, along with two normality tests: the Kolmogorov-Smirnov (KS) test and the Shapiro-Wilk test. Dr. Grande highlights the importance of interpreting these tests correctly, particularly when their results differ. He advises trusting the Shapiro-Wilk test when discrepancies arise and emphasizes the significance of normality in data analysis.

Takeaways

• 😀 Normality tests in SPSS help determine if a distribution is normal, which is important for data analysis in research.
• 😀 The script uses a typical SPSS data setup with 90 participants, two groups (control and CBT), and pre-test and post-test scores.
• 😀 To test for normality in SPSS, you can use the 'Explore' function found under the 'Analyze' menu.
• 😀 The dependent variables (pre-test and post-test scores) are added to the dependent list, and you can adjust plots and statistics settings in the dialog box.
• 😀 For testing normality, histograms and normality plots with tests are essential, and box plots can be omitted.
• 😀 Descriptive statistics such as skewness and kurtosis are important to evaluate the distribution of your data.
• 😀 The Kolmogorov-Smirnov (KS) test and Shapiro-Wilk test are used to test normality in SPSS.
• 😀 If the significance value from the normality test is greater than 0.05, you fail to reject the null hypothesis, suggesting the data is normally distributed.
• 😀 For the pre-test data, both KS and Shapiro-Wilk tests showed no significant differences, so the pre-test data is considered normally distributed.
• 😀 For the post-test data, both KS and Shapiro-Wilk tests showed significant differences, indicating the post-test data is not normally distributed.
• 😀 The Shapiro-Wilk test is often regarded as more reliable for testing normality than the Kolmogorov-Smirnov test.
• 😀 If KS and Shapiro-Wilk tests give different results, the Shapiro-Wilk test should be preferred for determining normality.
• 😀 Visual inspection of histograms can also help assess normality, with the pre-test histogram appearing normally distributed and the post-test one not.

Q & A

What is the purpose of testing for normality in SPSS?

The purpose of testing for normality in SPSS is to determine if the distribution of the data (e.g., pre-test and post-test scores) follows a normal distribution. This is important in research to ensure the appropriate statistical methods are applied.

What two normality tests are covered in this video?

The video covers two normality tests: the Kolmogorov-Smirnov (KS) test and the Shapiro-Wilk test. Both are used to assess if data significantly differs from a normal distribution.

Why does Dr. Grande recommend using the Shapiro-Wilk test over the KS test?

Dr. Grande recommends the Shapiro-Wilk test because it is generally considered more reliable than the KS test for assessing normality, especially with smaller sample sizes or specific types of data.

What does a significance value of less than 0.05 indicate in a normality test?

A significance value less than 0.05 indicates that the data is statistically significantly different from a normal distribution, leading to the rejection of the null hypothesis, which assumes the data is normally distributed.

What does a significance value greater than 0.05 indicate?

A significance value greater than 0.05 suggests that the data does not significantly differ from a normal distribution, leading to the acceptance of the null hypothesis, meaning the data is assumed to be normally distributed.

What role do histograms play in testing for normality?

Histograms provide a visual representation of the data distribution, helping to identify whether the data appears to follow a normal distribution. They complement the results from normality tests.

In this example, how were the pre-test and post-test data analyzed for normality?

In this example, both the pre-test and post-test data were analyzed by moving them into the 'Dependent List' in SPSS, selecting appropriate statistics and plots, and then interpreting the results from both the KS and Shapiro-Wilk normality tests.

What can be concluded if the Shapiro-Wilk test result for the pre-test is not statistically significant?

If the Shapiro-Wilk test result for the pre-test is not statistically significant (p > 0.05), it means the pre-test data is assumed to be normally distributed.

What can be concluded if both the KS and Shapiro-Wilk tests for the post-test are statistically significant?

If both the KS and Shapiro-Wilk tests for the post-test are statistically significant (p < 0.05), it indicates that the post-test data is not normally distributed.

Why might the Kolmogorov-Smirnov test and Shapiro-Wilk test sometimes yield different results?

The Kolmogorov-Smirnov test and Shapiro-Wilk test can sometimes yield different results because they analyze data in different ways. The Shapiro-Wilk test is often considered more reliable, particularly for smaller sample sizes or non-normal distributions.