In an EFA with 8 items, two factors were extracted. Five items have higher loadings on factor 1 and two items have more loadings on factor 2 in both the unrotated and promax-rotation matrices. However, there is this one item, which clearly had higher factor loading on factor 1 in unrotated matrix, demonstarted more loading on factor 2 with pro-max rotation. Why did this happen?
Unrotated factors are the first outcome from the underlying mechanics of factor analysis, but they are not necessarily the best representation of the data. Rotation examines whether there are other equivalent solutions may come closer to the standard of "simple structure," where each item loads on one and only factor.
In addition ProMax is an oblique rotation that allows for correlated factors, which is often a more realistic assumption.
If you have a two factor solution, I suggest that you examine the correlation matrix after you reorder it, so that you list the items from Factor One first, and then the items from Factor two. This "permuted" matrix will have two triangular pieces that contain the correlations within each of the factors, and a rectangular piece that contains the correlations between the two factors.
You should consider your theoretical model if it meets two correlated factors and remember that there are several methods of extraction and rotation
David L Morgan Cristian Ramos-Vera Thank you very much, Sir. The correlation between the factors was observed to be 0.48.
Can you please throw some light on differential item functioning in this context? Likelihood ratio test for DIF across gender categories was chosen as the data is ordinal in nature. Since we have a two factor solution, shall we perform this separately for both the factors?
Yes, it is, Sir. My understanding is that at a more fundamental level, if I would like to see gender variations in items, simple percentage differences between gender categories for each of the items would suffice. At the next level, when I am looking for item bias, I would consider the total score obtained across all the items, gender category, and the performance on each individual item.
Question: In this case where we are taking into account the total score obtained, what should the total score refer to, when I have a two factor solution? Should it refer to total score across all the items in the questionnaire or should it refer to total score obtained for items loaded on the factor on which the item loaded?
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix