I am hoping someone can help. I am running a CFA bifactor model and I have several items that are either not loading onto the general factor or are loading negatively onto the general factor. However, they are loading as hypothesized (and quite strongly I might add) onto the individual factors.
I would like to ultimately keep the items because 1. it strengthens the scale to have more items (there are only 17 items anyway), 2. because I think they are good items. I am struggling however because I cannot find any literature on having a general factor that is not a total score.
Any insight or literature references? Thanks so much in advance!