Sorry, is there any one who may have experience on PCA for a single factor extraction like the authors of the attached paper did for Trust and Loyalty with independent KMO value and share me please ? I read the paper and tried to do the same but, the PCA does not rotate for single factor.… Read more
Hi Zinashbizu
You haven't done anything wrong, it's just that you cannot rotate a solution with one factor. Rotation allows us to tell the software something about how the factors are related, basically whether they are correlated or uncorrelated. With one factor there is nothing to be correlated or uncorrelated with.
Cheers
Mark
Mark Shevlin , Thank you very much! I got your point but, I just wonder how could the authors of the attached paper could do that because they provide KMO value for each variable independently. And would you give me any suggestion regarding to how can I do factor analysis for IV, DV, and Mediation variable? I faced a problem of discriminant validity when I run factor analysis together for all these variables. Or could I take the mean value for dependent and mediation variable with out doing factor analysis please? Thank you again!
The most common goal in running factor analysis is to determine whether a set of items form a single scale. So, you would not combine the measures for your IV, DV, and Mediation variables into one factor analysis, because you are treating those as three different variables. Instead, you should conduct a separate factor analysis for each of them.
If each set of variables can indeed be combined into a single scale, then you can examine discriminant validity through the correlations among those three scales.
Dear David L Morgan , Thank you for your contribution. However, would you make me more clear please? I mean when I examine the correlation among three scales, as it is known that there will be strong correlation among dependent variable, independent variable and Mediator since they have cause and effect relationship. So, what will be my conclusion? or you mean Examine the correlation among the variables under IV,DV & Mediator independently just to see whether they are strongly correlated each other or not? The other question is that, can I use the mean value of DV and Mediator for Structural model or I have to threat them as a latent variable and include all the construct (questions) in Structural model please?
Thanks again!
Of course your variables should be correlated, but you mentioned a concern about discriminant validity, so you want to make sure that your variables are not so highly correlated (perhaps .8 and above) that they simply measure the same thing.
If you have multiple items for each scale, you do not have to use Structural Equation Modeling. Instead, you can sum the scores on the items for each scale (or compute a mean, whichever you prefer). There are technical reasons why using SEM will provide a better fit to the data, but that difference is often rather small in practice.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling