Canonical Correlation using R

Summary

This video module introduces Canonical Correlation as a method to test the relationship between two sets of continuous variables. It demonstrates how to use educational and entertainment-related responses from a survey to perform the test in RStudio. The process includes selecting the relevant variables, running the test, and interpreting results such as significance figures, F scores, Wilks’ Lambda, and canonical correlations. If significance is met, the results should be reported; otherwise, the test's failure to find a relationship should be noted. The video also offers guidance on output interpretation and reporting examples.

Takeaways

• 😀 Canonical Correlation is used to test if there is a relationship between two sets of continuous variables.
• 😀 A minimum of four continuous variables is required to perform the Canonical Correlation test.
• 😀 In the PEVS (Preferred Entertainment Viewing Survey) dataset, scale questions 3 through 7 are appropriate for continuous variables.
• 😀 The demonstration uses educational-related responses as one set of variables and entertainment-related responses as the other set.
• 😀 To run the Canonical Correlation test in RStudio, choose two sets of variables and input the relevant code in a Script box.
• 😀 You can copy the code directly from the Stat Tree module by pressing the [SCRIPT] button on the sending page.
• 😀 After running the test, the first result to check is the significance figure to determine if the relationship is statistically significant.
• 😀 If the significance figure meets your preset criterion, you need to report the test results in your research paper.
• 😀 When the test shows a significant relationship, report the test results, F scores, Wilks’ Lambda, squared canonical correlations, and canonical loadings.
• 😀 If the test does not show a significant relationship, report that no significant difference was found, and do not include test results in your paper.
• 😀 A full example of the output and how to report it can be found by pressing the [OUTPUT] button on the sending page of the Stat Tree module.

Q & A

What is the purpose of Canonical Correlation analysis?

Canonical Correlation analysis is used to test whether there is a relationship between two sets of continuous variables.

How many continuous variables are needed to run a Canonical Correlation test?

At least four continuous variables are required to run a Canonical Correlation test.

What are the two sets of variables used in the Preferred Entertainment Viewing Survey for this demonstration?

The two sets of variables used are educational-related responses for one set and entertainment-related responses for the other.

Which specific variables from the Preferred Entertainment Viewing Survey are selected for this demonstration?

The selected variables are 'history' and 'learn' for the educational-related responses, and 'entertain' and 'rewind' for the entertainment-related responses.

How is the Canonical Correlation test run in RStudio?

The test is run by creating a new Script box in RStudio, entering the appropriate code, and pressing the Run icon to generate output in the Console window.

What is the first result to check after running a Canonical Correlation test?

The first result to check is the significance figure associated with the test score.

What should you report if the significance figure meets the preset criterion?

If the significance figure meets the preset criterion, you should report the test results, including F score figures for any canonical correlations, Wilks' Lambda figure, squared canonical correlations for significant functions, and a table of standardized coefficients and canonical loadings for each variable.

What should you do if the significance figure does not meet the criterion?

If the significance figure does not meet the criterion, the test results do not need to be reported. However, you should report that the test failed to find a significant difference or relationship in the data.

Where can you find the output and an example of reporting the Canonical Correlation test results?

You can find the output and an example of reporting by pressing the [OUTPUT] button on the sending page of the Stat Tree module.

What are some key figures that need to be reported if a Canonical Correlation test is significant?

Key figures to report include F score figures for canonical correlations, Wilks' Lambda figure, squared canonical correlations for significant functions, and a table of standardized coefficients and canonical loadings.