No. The orthogonal setting means that the specific (group) factors are uncorrelated. The advantage of bi-factor EFA over Schmid-Leiman is discussed in Section 5 of the Jennrich $ Bentler (2011) article referred to in the Mplus Version 7 UG.
Eric Deemer posted on Saturday, April 06, 2013 - 1:12 pm
Okay, so the rotation is oblique between general and specific factors, but orthogonal across specific factors only?
I'm conducting an exploratory bifactor analysis. It looks like Mplus implements the Jennrich/Bentler approach. To handle potential problems with local minima, I want to follow the guidance of Mansolf and Reise (2017). They recommend at least 1,000 random starting values and examining at least the best 10 solutions.
I tried doing this in Mplus using the RSTARTS command ("RSTARTS 1000 10"), but the output included only a single solution. Is this because starting values all converged to the same solution? Or am I missing where I should be looking for the multiple solutions?
There is a section in the output called "Rotation function values at local minima, and start numbers:" To get multiple solutions you must have different numbers in the first column. If all the number are the same, the solutions are the same and we don't print these multiple times.