In my previous paper, I examined whether a personal Jensen coefficient could predict GPA beyond the general factor (or just the normal summed score). I found this not to be the case for a Dutch university student sample (n ≈ 300). One thing I did find, however, was that the personal Jensen coefficient was correlate with the g factor: r=.35.
Moreover, Piffer’s alternative metric, the g advantage coefficient (g factor score minus unit-weighted score) had a very strong correlation with the summed score r=.88. This measure is arguably thus a more reliable measure.
While neither of these predicted GPA beyond g, they may have another use. When there is teaching to the test, the subtests that increase the most are those that are the least g-loaded (see this). So, this should have an effect on these two measures, making them weaker or negative: the highest scores tending to be on the least g-loaded subtests. Thus, it may be practically useful to detect cheating on tests, although perhaps only at the group level.
Unfortunately, I don’t have any dataset with test-retest gains or direct training, but one could simulate gains that are negatively related to the g loadings, and then calculate the personal Jensen coefficient and Piffer’s g advantage coefficient.
Maybe I will update this post with the results of such a simulation.