Message/Author
 Bill T.P. posted on Friday, April 17, 2015 - 10:29 pm
In calculating the composite reliability, we start with summing all standardized coefficients for each factor. What if the signs of some of them are negative? Do we have to use the absolute values, or the native negative ones?
 Bengt O. Muthen posted on Saturday, April 18, 2015 - 7:52 am
I don't think you want to include an item that has a negative loading (unless the scoring was reverse).
 Bill T.P. posted on Saturday, April 18, 2015 - 8:16 am
As in exploratory factor analysis, the sign of an item's loading can be negative. I am sorry but I don't fully understand why I have to remove an item that has a negative loading.
 Bengt O. Muthen posted on Saturday, April 18, 2015 - 9:52 am
I am thinking that you want the reliability of a sum of items so that an increase in the sum reflects an increase in an underlying construct. Applying the omega formula you would sum loadings that are positive and negative which is not good. If you don't think a negative loading is substantively meaningful I would delete it, but if you think a negative loading is substantively meaningful, I would reverse the item so the loading is positive. Which I guess is the same as using the absolute value.