hi Frederique ,

maybe this article can help u interpret the data

Article Perception of employees toward e-learning service quality: E...

Thank You

Yes you have to also find their reliability with the help of cronbach aplha.

Hello Frederique,

There are a number of ways to compute factor/component scores (see this article for a very clear explanation of options and their virtues: http://pareonline.net/pdf/v14n20.pdf).

As well, you may wish to think about having used principal components extraction for your instrument. PCA tends to bias item-factor loadings upward compared to common factor analysis extractions, unless the number of variables/items is large (see Richard Gorsuch's 1983 text, Factor Analysis, 2nd ed., Lawrence Erlbaum).

Finally, as Mayanka Singh suggests, it is always a good idea to report estimated score reliability for your sample. That can help others who might consider using the instrument for their population of interest.

Good luck with your work!

Hi Frederique,

Your first question is actually two questions:

1a) What score to use following factor analysis, and

1b) How to deal with items that load unexpectedly.

I'm dividing this in two parts: a quick help and then some further suggestions.

Quick help:

1a) As David stated, you should calculate factor scores instead of simply averaging. If you use a simple average you could have saved you the trouble doing factor analysis at all.

I figure you used SPSS - you can easily store factor scores with this programm. There is the button "Scores..." where you can stick with the default "Regression". Remember turning this option off again or SPSS will clutter your data table with factor scores each time you run factor analysis (and one usually runs it multiple times with varying settings. A standard setting apart from SPSS's defaults would be: estimation method: maximum likelihood [refering to Davids comment]; rotation: promax).

1b) Yes, you should include PS1 (and getting the factor scores as described above will do so automatically).

2) Yes, you should calculate some measure of reliability.

Further suggestions:

1b) You seem to have a clear expectation which items should belong to which cluster. If you have such expectations, you should run confirmatory factor analysis (CFA), not exploratory factor analysis. CFA gives you means of testing the theoretically proposed model. Unfortunately, SPSS itself cannot do this, you would need other software (AMOS, MPlus, lavaan in R).

2) Consider not using Cronbachs Alpha, as it has some strong assumptions about your factor model. A good alternative would be McDonalds Omega. Again, SPSS will not give you this, unfortunately. But McDonalds Omega can be easily calculated from the factor loadings. If you would like to have an Excel sheet helping you with this just let me know here.