I am attempting to analyse data in SPSS version 17. I have 93 items dealing with ethical dilemmas faced and wish to work out if there are differences for those in rural versus city. Principal Components was suggested, and I have run this many times but I am concerned that the 'variables produced through this end up with residuals >3 and removal of these individuals, then rerunning the PCA means components change and there are 'new residuals'.
any help would be appreciated.
A lot of real data are skewed and it's necessary to check the data quality and do some transformations or collapsing. There are many methods to analyze this kind of data, factor analysis, logistic regression, even supervised/unsupervised clustering, it depends on your initial aim of the study.
Thankyou for responding. My concern is pretty much focussed on the issue that PCA etc will lead to the items where there was a lot of variance and possibly 'outliers' will be removed from the analysis. That said though, I want to ensure that my analysis is meaningful in the real world as well as statistically analysed in the most appropriate manner possible.
Hello Alvaro,
Thankyou for your response. The data was properly obtained, and the outliers are those who experienced more issues...hence are also what I am interested in.. that said though, the PCA did give 3 factors...and using the saved variables from the PCA, Anovas suggests differences for these across some independent variables such as location...
Was I right to use the saved regression variables from the PCA and use these in the Anova's? There seems to be a variety of ways individual create variables after running the PCA....
your help is really appreciated
What is the skewness of your original variables?. Maybe you should try to use some transformation to reduce skewness to assure that all data can be used into PCA properly. Let us know about the results.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS