I am wanting to model infant attachment as two latent constructs, anxiety and avoidance, using 4 reunion behavior rating scales (rated 1-9), from the "Strange Situation" measure of infant attachment, that have been shown to load onto 2 factors. There are no other 2-D infant attachment measures that I can incorporate into the methodology. Any suggestions as to how to produce 2 correlated indicators of each dimension from the 4 rating scales?
This is a difficult thing to do well given that you have only two indicators for each factor because a factor model with two indicators is not identified on its own. It is only identified in a model with another factor. This makes a two-factor model with two indicators for each factor sensitive to misspecification.