Wai Chow posted on Monday, November 28, 2005 - 2:09 pm
I’d like your advice on how to test model equivalence across reporters, i.e., child self-report (non-multi-level data; N=203) versus parent report of child competence (multi-level data, i.e., some parent pairs reporting on the same child; N=305; 191 moms, 114 dads; 203 clusters). Both versions of the measure has the near-identical 17 items, using a 4-point Likert scale (thus categorical data). Therefore, estimates should be computed using polychoric correlations. Hypothesized model across reporters is a bifactor model (i.e., three group factors and one general factor, all orthogonal to one another).
I have been instructed that since groups here are not independent (i.e., everyone is reporting on the same children, as opposed to groups determined by gender or ethnicity) errors should be correlated and equality constraints be imposed on factors and factor loadings. However, the issue of testing model equivalence when data of one reporter is multi-level and data of another reporter is not has us stumped.
Any advice you could offer on how to go about testing this very complex model would be appreciated. Thanks!
You could take a multivariate approach to modeling of families. It is described in the following paper:
Khoo, S.T. & Muthén, B. (2000). Longitudinal data on families: Growth modeling alternatives. Multivariate Applications in Substance use Research, J. Rose, L. Chassin, C. Presson & J. Sherman (eds.), Hillsdale, N.J.: Erlbaum, pp. 43-78
This application is a growth model but it can also be used in a cross-sectional setting. The trick is to treat each family as an observation where you have child variables, mother variables, father variables and family variables. If only one parent participates, you will have missing data.