| Educational measurement theorists disagree about which approach for assessing dimensionality is most appropriate for a dichotomously-scored test. In this study, five statistical approaches for assessing dimensionality are applied to the content portion of the 1997 School Achievement Indicators Program (SAIP) Mathematics Assessment. The five approaches are: (1) essential dimensionality, (2) confirmatory linear factor analysis, (3) exploratory linear factor analysis, (4) limited-information nonlinear factor analysis, and (5) full-information nonlinear factor analysis. The results suggest that nonlinear full- and limited-information approaches are the most appropriate methods for evaluating the dimensionality of the SAIP Mathematics Assessment and similar dichotomously-scored data sets. A brief summary of the benefits and drawbacks of each of the computer programs used to implement the theoretical approaches is provided. |
