# Interpreting negative correlations between factors?

#### capybara

##### New Member
Hello evetyone!

I conducted factor analysis on a 30-item scale which is suppose to be measuring xenophobia. All the variables were previously recoded in order to express the same side of the measured dimension. (I am using SPSS)

I intend to analyse the relationship between xenophobia and a couple of predictors. Instead of doing multiple regression with the sum score from the scale as the criterion, I intend to use the extracted factors as criterions in canonical correlation analysis. But, before that, I need to understand a couple of things concerning FA...

I did FA with oblique rotation (Oblimin) and maximum likelihood as the extraction method, and chose to keep the 3 factor solution. The second factor has mainly negative loadings and therefore correlates negatively with he other two factors. I would expect the loadings and correlations to be positive since all the items were recoded, so I have trouble interpreting the negative correlations and understanding why they happen.

I hope someone more experienced in factor anlysis can help me out. Thanks!

#### spunky

##### Doesn't actually exist
I intend to use the extracted factors as criterions in canonical correlation analysis.
Care to elaborate a little bit more on this? Because of the problem of factor indeterminacy, it's generally not a good idea to extract factor scores and do subsequent analyses on them. But I might be misunderstanding you and you're actually planning on doing something else.

I did FA with oblique rotation (Oblimin) and maximum likelihood as the extraction method, and chose to keep the 3 factor solution. The second factor has mainly negative loadings and therefore correlates negatively with he other two factors. I would expect the loadings and correlations to be positive since all the items were recoded, so I have trouble interpreting the negative correlations and understanding why they happen.
How did you choose a 3 factor solution? Is your data ordinal/categorical in nature (it seems like it is)? What's your sample size?

I hope someone more experienced in factor anlysis can help me out. Thanks!