

HigherOrder Factor Analysis 

Message/Author 

df_txst posted on Sunday, February 18, 2018  7:06 pm



I conducted an EFA in SPSS and found a fourfactor solution: shape = 2 factors, color = 1 factor, size = 1 factor. The 2 factors that explain shape perfectly load separately as regular and reversecoded items. When I calculate the to scales separate, the correlation between them is low and each of their correlations with the combined scale is high. Good right? So I'm assuming if I had the two scales in the model and specified them to load on a higherorder SHAPE latent factor, then that loading would work similar. Is this the best approach to use? and If so, how exactly to is specify this in Mplus? Btw: I'm ultimately trying to run a path analysis from shape to color to size. 


I am not sure why you would specify a higherorder factor behind the two scales if they have a low correlation  factors are supposed to capture correlations. Chapter 5 in the UG shows how to specify higherorder factor analysis. With only 2 firstorder factors, their loadings need to both be fixed. 

Back to top 

