The G coefficient and its relation to R-type factor analysis

The G coefficient is proven to be an inaccurate indicator of latent relations and, therefore, unacceptable as a measure of association when used in R-type factor analysis. On the other hand, the G coefficient is useful in measuring another type of relation, the simple agreement of scores for a pair of items. If one thinks of the scores for each item as a vector, the G coefficient is functionally related to the Euclidean distance between the endpoints of the two vectors.