SPSS Correlation Analysis
Correlation analysis is used to study the relationships between quantitative data, including whether relationships exist and how closely they are related. This analysis method is typically used before regression analysis. The logical relationship between correlation analysis and regression analysis is: there must be a correlation relationship first before there can be a regression relationship.
What is Correlation Analysis
Correlation analysis evaluates the relationships between quantitative variables. It uses correlation coefficients to represent relationships between analysis items. The analysis process involves three steps:
1. Determine if a relationship exists
2. Determine the direction of the relationship: A correlation coefficient greater than 0 indicates a positive correlation, while a value less than 0 indicates a negative correlation.
3. Determine the strength of the relationship: Generally, a correlation coefficient greater than 0.4 indicates a close relationship.
Important Note: Correlation analysis is typically performed before regression analysis. If there is no correlation relationship, there cannot be a regression relationship. However, having a correlation relationship does not necessarily mean there will be a regression relationship.
Correlation Coefficient Types
The system provides three types of correlation coefficients: Pearson, Spearman, and Kendall. The system defaults to using Pearson correlation coefficient. The differences between these coefficients are as follows:
| Coefficient | Usage Scenario | Notes |
|---|---|---|
| Pearson | Quantitative data that satisfies normality | Normality can be checked using PP/QQ plots, histograms, or normality tests (most strict test) |
| Spearman | Quantitative data that does not satisfy normality | Normality can be checked using PP/QQ plots, histograms, or normality tests (most strict test) |
| Kendall | Quantitative data for consistency judgment | Usually used for rating data consistency level research (not relationship research), such as judge scoring, data ranking, etc. Kendall tub_b type |
Note: In theory, if data is normally distributed, use Pearson correlation coefficient; if data is not normally distributed, use Spearman correlation coefficient. However, normal distribution rarely exists in practice. As long as the non-normal situation is within acceptable limits, Pearson coefficient can continue to be used. In most cases, conclusions from Pearson and Spearman coefficients remain basically consistent. Therefore, the vast majority of research uses Pearson correlation coefficient, while Spearman is used less frequently.
Feature Access
1. Navigate to the "Analyze Results" section of your questionnaire in the SurveyMars system.
2. Click on the "SPSS Analysis" option to access the analysis features.
3. Click on the "Analysis now" button and select "Correlation Analysis" from the available analysis methods.
Performing Correlation Analysis
1. Select the quantitative variables you want to analyze for correlation.
2. From an analysis method perspective, correlation analysis does not distinguish between X and Y. However, from an analysis logic perspective, it is recommended to distinguish X and Y (e.g., if studying how "Service Attitude" and "Service Quality" relate to "Satisfaction" and "Loyalty", the first two are X and the latter two are Y).
3. Click the "Confirm" button to generate correlation analysis results.
Interpreting Correlation Results
Correlation analysis results are interpreted in three steps:
1. Determine if a relationship exists:
- If there is an asterisk (*) in the result, it indicates a significant relationship exists
- One asterisk (*) indicates significance at the 0.05 level (p < 0.05)
- Two asterisks (**) indicate significance at the 0.01 level (p < 0.01)
- If there is no asterisk, it indicates no significant relationship
2. Determine the direction of the relationship:
- Correlation coefficient > 0: Positive correlation (as one variable increases, the other also increases)
- Correlation coefficient < 0: Negative correlation (as one variable increases, the other decreases)
3. Determine the strength of the relationship:
- Correlation coefficient > 0.7: Very close relationship
- Correlation coefficient 0.4 - 0.7: Close relationship
- Correlation coefficient 0.2 - 0.4: Moderate relationship
- Correlation coefficient < 0.2: Weak relationship (but still significant if marked with asterisk)
Example Result Interpretation:
If "Online Shopping Satisfaction" and "Repurchase Intention" show a correlation coefficient of 0.673** with p < 0.01, this indicates:
- There is a significant relationship (indicated by **)
- The relationship is positive (coefficient > 0)
- The relationship is close (coefficient between 0.4 and 0.7)
Important Notes
- Correlation analysis is used to study relationships between quantitative data only
- Correlation analysis is typically performed before regression analysis; there must be a correlation relationship before there can be a regression relationship
- The system defaults to using Pearson correlation coefficient, which is suitable for most research scenarios
- Correlation analysis does not distinguish between X and Y from a method perspective, but it is recommended to distinguish them from an analysis logic perspective
- A correlation coefficient greater than 0.4 generally indicates a close relationship
Frequently Asked Questions (FAQs)
Q1: Where is the p-value in the results?
A: The p-value (also called significance value or Sig value) is represented by asterisks in the correlation coefficient table. For correlation analysis, the standard table format uses asterisks to indicate p-values (marked in the upper right corner of correlation coefficients). p < 0.01 is represented by 2 asterisks (**), and p < 0.05 is represented by 1 asterisk (*).
Q2: How should I handle multiple scale items that represent one dimension?
A: If multiple scale items represent one dimension (e.g., two items both representing "Loyalty"), you can use the "Generate Variable" function with the "Average" option to combine multiple scale items into one overall dimension. After combining items into a whole, you can then perform correlation analysis, regression analysis, variance analysis, etc. using the whole dimension rather than individual items.
Q3: What should I do if data is not normally distributed?
A: Theoretically, if data is normally distributed, use Pearson correlation coefficient; if data is not normally distributed, use Spearman correlation coefficient. However, normal distribution rarely exists in practice. As long as the non-normal situation is within acceptable limits, you can continue using Pearson coefficient. In most cases, conclusions from Pearson and Spearman coefficients remain basically consistent. Therefore, the vast majority of research uses Pearson correlation coefficient, while Spearman is used less frequently.
Q4: Should I distinguish between X and Y variables?
A: From an analysis method perspective, correlation analysis does not distinguish between X and Y. However, from an analysis logic perspective, it is recommended to distinguish X and Y. For example, if studying how "Service Attitude" and "Service Quality" relate to "Satisfaction" and "Loyalty", the first two are X (independent variables) and the latter two are Y (dependent variables). If you do not want to distinguish X and Y, you can place all items in the "Analysis Items Y (Quantitative)" box.
