Estimation Methods for EFA in JASP
Dear all,
I have been struggling to understand the computational differences between the different estimation methods that JASP uses for EFA when comparing them to the "psych" package in R.
Essentially, I am trying to use an estimation method that is suited best for slight multivariate normality violation and some heteroscedasticity in the data. I thought "minres" is the best solution here because according to the psych package author it is better to use "Ordinary Least Squares (OLS) to find the minimum residual (minres) solution" (https://www.rdocumentation.org/packages/psych/versions/1.8.12/topics/fa).
However, in JASP, I can choose MINRES or Ordinary least squares separately, so I was wondering what the difference is?
Beyond that, wouldn't generalized least squares be potentially superior to minres for non-normal data (or at least weighted least squares)?
Any advice is appreciated!
Comments
From the documentation at your link:
fm="ols" differs very slightly from "minres" in that it minimizes the entire residual matrix using an OLS procedure but uses the empirical first derivative. This will be slower.
and a bit later:
Following extensive correspondence with Hao Wu and Mikko Ronkko, in April, 2017 the derivative of the minres and uls) fitting was modified. This leads to slightly smaller residuals (appropriately enough for a method claiming to minimize them) than the prior procedure. For consistency with prior analyses, "old.min" was added to give these slightly larger residuals. The differences between old.min and the newer "minres" and "ols" solutions are at the third to fourth decimal, but none the less, are worth noting. For comparison purposes, the fm="ols" uses empirical first derivatives, while uls and minres use equation based first derivatives. The results seem to be identical, but the minres and uls solutions require fewer iterations for larger problems and are faster. Thanks to Hao Wu for some very thoughtful help.
In short, it will be extremely similar. We provide this option only because psych provides it.
As to your last question, I don't have a good answer ready. There might be some simulation studies out there comparing the different estimation methods for non-normal data.
Hope this helps!
Erik-Jan
Thank you, this is helpful!