factoran() vs EM Algorithm FA = fa() : Understanding the difference on carbig.mat in ...toolbox​\stats\sta​tsdemos

Hi,

Could someone explain me the difference between Matlab built-in

 factoran(x,k)

vs user-defined

 EM algorithm fa(x,k)

-------------------------------?? ( code )

load carbig;
X = [Acceleration Displacement Horsepower MPG Weight];
X = X(all(~isnan(X),2),:);
lambda= factoran(X, 2)

and

lambda2=fa(X,2) % from <http://www.mathworks.com/matlabcentral/fileexchange/28906-factor-analysis>

Specically I'd like to know why

factoran

outputs a (5*2) matrix of loadings with

size(X,1) = 120

whereas:

EM algorithm fa

outputs a (120*2) matrix of loadings matching my expectations, to the extent that for each data/variable i end up with a sensitivity of the factors.

Best,

Answer 1

[Tom Lane]

It sounds like you want the fifth output of factoran rather than the first.

[Isma]

That's what i've realized ex-post . it sounds like the EM algo outputs the factor 'scores' (not the factor loadings), matching more or less the 5th output of factoran. cheers.