SPSS
1. Reliability analysis
2. Evaluating items
3. Rules for deleting an item
4. Compute all individual total scores
5. Recode items
6. Principal component analysis (PCA)
7. KMO and Bartlett’s test of sphericity
8. What is the commonality?
9. What is the eigenvalue?
10. Component loading
11. Choosing components based on the Kaiser-Guttman rule
12. Choosing components based on the scree plot
13. Run the analysis with a pre-specified number of components and a rotation
14. Violation of a simple structure
15. Which items belong together
16. Compute scale scores
17. Correlation between items and components
18. Procedure to construct scales
19. Compute KMO
20. Choosing rotation method
21. Why choose a Varimax solution?
22. Which item should be recoded
23. Name the items and their patterns
, 1. Reliability analysis
Go to Analyze > Scale > Reliability Analysis and drag all items of the tolerance scale (v225
through v242) into the Items window (Right-click in the Items window and choose “Display
variable names” to see the names of the items).
Then click on the Statistics button and enable the following options before you paste the
syntax:
- Item, Scale and Scale if item deleted in Descriptives for
- Correlations in Inter-Item
Click on Continue, and paste and run the syntax.
the reliability is the Cronbach’s alpha
Example:
- Which item contributes the most to the reliability of the scale? Why?
- The item having casual sex contributes the most to the reliability of the scale. Of all
items, it has the highest corrected item-total correlation and the lowest Cronbach’s
alpha if item deleted.
2. Evaluating items
- To evaluate items, we use the following rules-of-thumb:
o an item-total correlation of at least 0.3
o Cronbach’s alpha if deleted should be lower than Cronbach’s alpha of the
whole scale
- if an item does not meet one of these criteria, it requires closer inspection.
- In principle, if we are being strict, all items contribute positively to the reliability of
the scale: all corrected item-total correlations are > 0. However, if we look in more
detail, we can see a few items with striking properties/values – see next slide.
Example:
- One item should be deleted from the scale, which one?
Concerning Q80E, it is very concerning that the item-total correlation is really low, since this
can mean that this item does not measure the same thing as the other items (since there is
basically no correlation between the answers of respondents to this item and the respondents
answers to all other remaining items). This might be an indication that this item does not
belong in this scale (for example content-wise) or that respondents did not understand the
item correctly or their answers are biased (e.g., it could be an item that triggers socially
desirable answers). Based on solely the results of the reliability analysis, we would thus
decide to delete this item.
1. Reliability analysis
2. Evaluating items
3. Rules for deleting an item
4. Compute all individual total scores
5. Recode items
6. Principal component analysis (PCA)
7. KMO and Bartlett’s test of sphericity
8. What is the commonality?
9. What is the eigenvalue?
10. Component loading
11. Choosing components based on the Kaiser-Guttman rule
12. Choosing components based on the scree plot
13. Run the analysis with a pre-specified number of components and a rotation
14. Violation of a simple structure
15. Which items belong together
16. Compute scale scores
17. Correlation between items and components
18. Procedure to construct scales
19. Compute KMO
20. Choosing rotation method
21. Why choose a Varimax solution?
22. Which item should be recoded
23. Name the items and their patterns
, 1. Reliability analysis
Go to Analyze > Scale > Reliability Analysis and drag all items of the tolerance scale (v225
through v242) into the Items window (Right-click in the Items window and choose “Display
variable names” to see the names of the items).
Then click on the Statistics button and enable the following options before you paste the
syntax:
- Item, Scale and Scale if item deleted in Descriptives for
- Correlations in Inter-Item
Click on Continue, and paste and run the syntax.
the reliability is the Cronbach’s alpha
Example:
- Which item contributes the most to the reliability of the scale? Why?
- The item having casual sex contributes the most to the reliability of the scale. Of all
items, it has the highest corrected item-total correlation and the lowest Cronbach’s
alpha if item deleted.
2. Evaluating items
- To evaluate items, we use the following rules-of-thumb:
o an item-total correlation of at least 0.3
o Cronbach’s alpha if deleted should be lower than Cronbach’s alpha of the
whole scale
- if an item does not meet one of these criteria, it requires closer inspection.
- In principle, if we are being strict, all items contribute positively to the reliability of
the scale: all corrected item-total correlations are > 0. However, if we look in more
detail, we can see a few items with striking properties/values – see next slide.
Example:
- One item should be deleted from the scale, which one?
Concerning Q80E, it is very concerning that the item-total correlation is really low, since this
can mean that this item does not measure the same thing as the other items (since there is
basically no correlation between the answers of respondents to this item and the respondents
answers to all other remaining items). This might be an indication that this item does not
belong in this scale (for example content-wise) or that respondents did not understand the
item correctly or their answers are biased (e.g., it could be an item that triggers socially
desirable answers). Based on solely the results of the reliability analysis, we would thus
decide to delete this item.
