confused about factor analysis output

#1
Hi,

I have 100 participants answering an aesthetic scale ( 8 items in the scale) in my data. The question is "rate the following statements from 1= strongly disagree to 7= strongly agree". I copied 2 statements from my scale below.

1/Owning products that have superior designs makes me feel good about myself
2/I enjoy seeing displays of products that have superior designs


I have never done a factor analysis before but I learned the steps in class. However, my results are funny.

Analyze - Dimension Reduction - Factor.
Included 8 items in the scale.
Extraction using the principles component method.

My problem is when I conduct a factor analysis I only get 1 component which has an eigen value bigger than 1. Therefore, no rotations etc. I looked at several examples online and also an in-class example (12 items scale) that got more than at least 3. I am feeling there is something wrong but not sure what it is. Can we get 1 component in Factor Analysis. If yes, what does it mean? if no, what am I doing wrong, how do I fix it?

I hope someone can explain this to me. Really need some help.

Thanks so much for your time,
Ege
 
#2
Hi Ege,

It's totally fine to have only one component, especially with only 8 items. Do all 8 items load on it (loadings > .4)?

It just means that all 8 items are measuring a single construct. Does that make sense in this context or were you expecting more constructs?

However, the eigenvalue>1 criterion is not a good cutoff. You can force the model to contain 2 components and see how the results differ. http://www.theanalysisfactor.com/factor-analysis-how-many-factors/

Best,
Karen
 
#3
I agree with the statement above. If the phenomenon in question has a simple latent structure it does. Be happy....:p Commonly you don't do EFA with just 8 measure, 30 or more I think are recommended and it is often far more. So examples will tend to generate multiple factors.
 
#5
It should be noted that in common usage, using principles component method to extract factors which are then analzyed with rotation is called EFA even though it is not. This is in fact the default extraction method for SAS's exploratory factor analysis software. Principal component analysis actually removes 100 percent of the variance (keeps pulling factors until you have 100 percent) so formally I don't think you are doing that if you (as is common with "factor analysis") set some rule on how many factors you extract such as using a scree plot).
 

Karabiner

TS Contributor
#6
It should be noted that in common usage, using principles component method to extract factors which are then analzyed with rotation is called EFA even though it is not.
Yes, unfortunately, since it leads to some confusion, and sometimes even
to wrong conclusions. But during the last few years the more appropriate
usage of terms seems to have gained ground, so hope is not in vain...

With kind regards

K.
 
#7
I don't karabiner - my EFA class two years ago called methods that use PCA as an extraction took EFA througout. It was only in reading articles on factor analysis in journals this year that I realized this issue even exists. That SAS still uses PCA as an extraction tool, given how often it is used in business, suggests limited progress outside academic circles.

I had not realized this was SPSS forum. SPSS does use a factor analysis extraction method as its default in its EFA tool.