We make no pretense that the procedures we have described are exhaustive: There may be other creative ways to link measures together across studies and additional research on this topic is needed. Nor do we assume these procedures will be universally applicable. For instance, in the models illustrated here we assumed that the items reflect a continuous underlying latent variable and that, conditional on this latent variable, the item responses were locally independent. These assumptions would not be appropriate if, for instance, one wished to score a categorical latent variable (e.g., diagnosis) or if item responses were nested within reporters (e.g., parent, teacher, and self reports). Other psychometric models would be needed in such circumstances. Nevertheless, we believe that the general set of steps illustrated here for constructing commensurate measures across studies should generalize to a wide variety of measurement scenarios.
