Exploiting redundant measurement of dose and developmental outcome: New methods from the behavioral teratology of alcohol.

Abstract

In human teratology, the magnitude of exposures cannot be experimentally controlled. But when dose and response are both measured redundantly, a new statistical approach, Partial Least Squares, can powerfully summarize the data while avoiding familiar inferential pitfalls. Procedures of this class have been exploited in a study of the enduring effects of prenatal alcohol exposure upon neurobehavioral development of some 500 children born in 1975-1976. In the context ofthat study, this article reviews the new methodology under 4 headings: (a) descriptions of dose and outcome measures jointly by way of new composite scores and their saliences (cross-correlations), (b) covariate adjustments for these summaries and scores, (c) methods for inspecting the dose-response relationship between the composites and the original measures, and (d ) methods for studying the emergence or amelioration of deficits across developmental age.