Application of structural modeling with latent variables to adolescent drug use: a reply to Huba, Wingard, and Bentler.