Hello, I have first cleaned the data by handling all missing values.
1. I have done factor analysis for four variables regarding environmental concern.
2. I used PCA and Varimax
3. I have got only one component with an engine value greater than 1.
4. So the component cannot be obviously rotated
5. I tried changing the method of factor analysis to Maximum likelihood and Direct oblimin as I did not want to limit my data to orthogonal rotation.
6. I got the same results with components count and the chi-square test has a significance of .000
7. The correlation matrix table for all the four variables also have either .000 or
With just four variables, there is not much room for extracting more than 1 factor. Nonetheless, the fact that the chi-square test of model fit is significant may indicate that the four variables are not unidimensional. Without seeing your correlation matrix and other details of your data and analysis, it is impossible to say.
I dont want to be picky, but PCA is not a factor analysis! Just in case you want to present/publish your results and/or you initially intendend to do a factor analysis you should be clear what you did or wanted to do.
Hello Shivali,
The first point to address is, what is your goal here, and how does principal components or common factor analysis help you to reach that goal?
What I can say, given your query, is:
1. The ML result (chi-square with p-value < .001) implies that there is more information in the data set (of four variables) than is captured by one factor.
2. I'm not sure what you mean concerning the correlation matrix, unless it is that the correlations for all pairs of observed variables are significantly different from zero.
3. In factor analysis, using the eigenvalue > 1 criterion for deciding how many factors to extract is probably one of the least satisfactory methods available. Parallel analysis is better.
4. And yes, as Rainer Duesing points out, PCA is not common factor analysis: It makes strong assumptions (all observed variance is common variance, there is no error or specific variance in the data set) which, in my experience, are seldom correct.
Good luck with your work.
- Badges
- Science topic