Message/Author 

Li Lin posted on Thursday, February 18, 2010  10:45 am



I am repeating an EFA using the same data and the same Mplus code, but in different Mplus versions. The output were identical, except some factors had its factor loadings sign flipped(absolute values didn't change). What caused this change in the sign? Would this be a problem for using the EFA analysis result? Please help. Thanks! Part of the output: Mplus VERSION 5  QUARTIMIN ROTATED LOADINGS 1 2 ________ ________ SFGLO01 0.419 0.444 SFDES01 0.895 0.028 SFDES02 0.909 0.133 SFDES03 0.866 0.345 SFDES04 0.721 0.193 RECDES05 0.597 0.419 RECDES06 0.024 0.555 RECDES07 0.088 0.727 SFDES08 0.217 0.413 ... Mplus VERSION 5.21  QUARTIMIN ROTATED LOADINGS 1 2 ________ ________ SFGLO01 0.419 0.444 SFDES01 0.895 0.028 SFDES02 0.909 0.133 SFDES03 0.866 0.345 SFDES04 0.721 0.193 RECDES05 0.597 0.419 RECDES06 0.024 0.555 RECDES07 0.088 0.727 SFDES08 0.217 0.413 ... 


If all of the signs in a column change, this is not a problem. There was probably a slight algorithm change between Version 5 and 5.21. It just changes the factor meaning from, for example, inability to ability. 

Chie Kotake posted on Wednesday, April 30, 2014  2:03 pm



Hi! I have a similar situation, but I'm using the same version to test 2 models. Basically, the first model is correlating two variables (A<>B). The second model uses the same model, except now it's A predicting B (A>B). The first model has positive factor loadings, but when I switch to the second model, the factor loading sign and the means signs flips. Do you have any suggestion as to why this is occurring? I'm using version 7.11. Thank you so much for your help! 


It just finds the opposite solution. This is not a problem. Use positive starting values for the one that switched. 


Hello! I'm doing a singledimensional factor analysis of categorical items where I'd like to obtain flipped loadings. I tried to fix the loading of the first indicator to 1 rather than 1, but wasn't able to deduce the correct syntax for this. MODEL: f1 by y1y5; 


f1 by y1@1 y2y5*1; 


Tack så mycket! (Thank you very much!) 

Back to top 