Social-indicator model building: A multiple-indicator design

The purpose of the paper is to illustrate and apply a technique for building social-indicator models when concepts are measured by multiple indicators. The multiple-indicator approach discussed in the paper is based on the work of John L. Sullivan in which multiple-partial correlation is used in block-recursive model building. Sullivan's approach is based on the earlier work of Hubert M. Blalock with partial correlation. The multiple-indicator approach is initially considered by explaining the function of multiple-partial correlation and illustrating the utility of the statistic in sociological analysis. The model building approach is first illustrated in the case of a four-variable model measured by single indicators. An elaboration of the design is presented in the case of a six-variable model in which the variables are measured by multiple indicators (two indicators per variable). The multiple-indicator approach is then applied to assess the goodness-of-fit between data and a macrosociological health model. Procedures for assessing fit are presented and discussed. Finally, several advantages and disadvantages of the multiple-indicator approach to model building are considered.