I would like to know if its possible to compare the factor loadings obtained from a particular survey with mine using the same questionnaire. This is to enable me to establish the variables capable of explaining the observed variance in my study and to confirm if there are differences in the item loadings.
Dear Olubusayo
It would be useful to know more details about what you are intending to do. But, generally speaking, comparing loadings values of two Principal Component Analysis does not seem very useful to me. Instead I would analyze:
1. If the factors of the second PCA are identical to the first one (If they have the same meaning).
2. If there is variation of the explained variability of each factor and of the whole model (which could me that some factors are more important in a given population)
3. If any particular item disappeared from a factor (because de loading was
Dear Dr Filipe,
I appreciate your prompt response to my mail as I was glued to my computer hoping to get an answer as soon as possible. I wanted to understand the attitude of a given population and I adopted a particular questionnaire. Then i realized the attitude in my population is quiet the opposite. This might have been because of the orientation and culture of my study population. Conducting the PCA for five factors;
1) No item disappeared from the factor because the loading was greater than 0.4
2) Two factors were identical to the first PCA from the initial study.
3) Three other factors were somewhat identical but they had negative signs. Does the negative sign obtained for my PCA means that my sample had negative attitude in these factors?
4) Can the PCA that I conducted also establish content and factorial validity of the questionnaire because of its relevance to my study population
Thank you.
The strongest argument that you could make would be to do confirmatory factor analysis, as opposed to "eyeballing" whether one exploratory factor analysis (PCA) "looks like" another.
This would mean setting up the prior factor structure as a hypothesis: that you would apply to your data, and then comparing the fit of that structure to the one you obtained through your analysis.
Concerning your last questions, it is good sign that no item disappeared from your PCA, that means that they are relevant to the components you found.
There is a good explanation for the negative values your are getting here:
https://www.researchgate.net/post/What_is_the_meaning_of_negative_values_in_components_from_PCA_analysis
If your PCA got a model for your population that you can interpret and the factors are relevant for what your are studying, then yes, the questionnaire shows factorial validity for your population, even if the factor structure is different from the original study.
I also want to share that when I started to use factorial analysis, this book was very useful to me:
http://books.google.pt/books/about/Making_Sense_of_Factor_Analysis.html?id=9kB5jE2IjS4C&redir_esc=y
Thank you Dr Filipe for being so explicit, I found the links very useful and I feel more confident to continue with my analysis. I also appreciate Prof David, it's a privilege that you look through my questions. Thank you.
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology