ABSTRACT
Statistical methods for handling ordinal categorical variables in social and behavioral research have gained considerable attention in recent years. In analyses of categorical variables, thresholds are provided for each increasing scale level, that is, each threshold represents a portion of the underlying continuous scale. Because ordered categorical variables do not have origins of measurement, the means and covariances of the underlying continuous variables are not identified. In EQS, structural equation modeling (SEM) with ordered categorical variables can be accommodated using a modification of the Lee-Poon-Bentler approach. Despite the recent advances in handling ordered categorical variables and the ubiquitous nature of nonnormally distributed data in the social and behavioral sciences, applications of these methods are not widespread. This chapter provides only a brief introduction to growth modeling with ordered categorical outcomes. A complete exposition of the various methods employed within current SEM programs is beyond the scope of the chapter.
