No abstract

Chapter 2 is a startling reminder that in longitudinal research, no data analysis approach, however well established or widely used, can be assumed to be fully effective in a particular setting. Instead, careful scrutiny must be given to every application. Cohen gives the reader who uses covariatespractically every reader who analyzes empirical data-plenty to scrutinize. This chapter raises the problem of the "premature covariate," which arises when a covariate is changing over time. Cohen shows that, quite apart from any measurement error considerations, if a covariate is not measured at the time it exerts its causal influence, it may not be an effective covariate; that is, the statistical analysis may not partial out all of the effect of the covariate on the dependent variable. The consequences can be substantially biased estimates of effects. The argument and example given in this chapter involve reciprocal causation, but the general idea holds whenever a covariate or dependent variable is dynamic.Two comment sections follow this chapter. In the first, Collins and Graham outline the general problem of timing of measurements as a design consideration. They argue that this issue is relevant to most large-scale longitudinal studies. In the second comment section, Little frames this as a missing data problem, where the covariate at the time it exerts its effects constitutes an array of missing data. Together, chapter 2 and the two com-18

No abstract

No abstract