Latent Variable Techniques for Measuring Development

Abstract

The seminal work of Lazarsfeld (1950a, b) on latent structure analysis carried out more than three decades ago charted a new direction for research involving relationships between latent and manifest variables. The two branches of investigation stemming from Lazarsfeld’s early work, the latent class branch and the latent trait branch, can be used to address a variety of problems of importance in measuring development. This chapter examines latent class and latent trait models in the context of a large-scale application involving the measurement of development in children participating in the Head Start program. The chapter reviews the application of latent class models describing the ordering of skills in a developmental sequence, and it examines the question of determining the effects of one skill on another. Then, the application of latent trait models for testing hypotheses about developmental sequences is discussed. Hierarchical models are reviewed that constrain slope and difficulty item parameters to test hypotheses about difficulty ordering and slope uniformity for item sets reflecting developmental sequences.