I used the transformational (TFL) and transactional (TAC) scales of De Hoogh, Den Hartog and Koopman (2004) and the items for both leadership styles load in one factor in the factor analyses (eigenvalue > 1). Only when I analyse a fixed number of 2 factors, most of the factors load well in factor one or two. What does this means and can I treat them as separate scales or not?
The evidence for two scales is not strong. About 10 items in your two-factor solution load on both factors. I'd suggest moving forward with the one factor solution. The specific leadership situation may have affected the transactional/transformational breakdown. Good luck with your study...Nick
Dear Wilfred,
I agree with Nicholas regarding your two component solution. However, there are other options. I would recommend Principal Axis Factoring rather than Principal Component Analysis.
The rule we recommend is to remove items which cross load with a ratio > 4:3. By removing the items with the highest cross loading ratio in turn you should end up with a stable solution with two factors. These factors can then be validated using Principal Component Analysis and Cronbach's alpha as a cross check.
Before deciding that dataset has a two-factor solution or single-factor, there is a number of things you need understand. For instance:
- What is the sample size; you need to ensure 1:10 (item: Respondents) to get authentic results
- If you are doing EFA then use Principal Component Analysis otherwise you need to opt Principal Axis Factoring rather than.
- Check KMO; if it is 1
- Use rotation Varimax
- Exclude those items with cross factor loading >.4; there is issue of Multicollinearity
- While running factor analysis in SPSS, press “option” button and check both ‘sorted by size’ and ‘suppress small coefficient’ to make output more readable
- Ensure there is no item is reversed coded.
When you delete an item, in addition to statistical justification, you must have theoretical support as well to explain the reason why you have deleted an item.
In spite of following these steps, if you still to able to get correct output, then look closely the data you enter. Find those responses in which respondents carelessly responded with swing tendency (high to low and low to high). Delete those responses and run factor analysis again.
Lastly, if you are still not able to get a two-factor solution or getting a two-factor solution with mix items then the only issue remains relevant is either the selection of questionnaire is inappropriate or the target population is not able to differentiate two kinds of leadership styles.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis