Premium
Prospective studies of exposure to an environmental contaminant: The challenge of hypothesis testing in a multivariate correlational context
Author(s) -
Jacobson Joseph L.,
Jacobson Sandra W.
Publication year - 2004
Publication title -
psychology in the schools
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.738
H-Index - 75
eISSN - 1520-6807
pISSN - 0033-3085
DOI - 10.1002/pits.20002
Subject(s) - psychology , bonferroni correction , context (archaeology) , multivariate statistics , confounding , test (biology) , assertion , developmental psychology , social psychology , cognitive psychology , statistics , computer science , mathematics , paleontology , biology , programming language
In this paper, we respond to the criticisms and concerns raised by D.V. Cicchetti, A.S. Kaufman, & S.S. Sparrow (this issue) in their review of the PCB literature, with particular attention to our own research in Michigan. We agree that multiple comparisons and functional significance are issues that would benefit from more discussion. However, because the effects associated with exposure to environmental contaminants are generally subtle, the risk of Type II error would be unacceptably high if researchers were to adopt the authors' recommendation to use a Bonferroni correction. We describe the hierarchical approach we have used to deal with the issue of multiple comparisons, which emphasizes the need to base interpretation on consistent patterns in the data and on replicated findings. The issue of confounding is one that has received considerable attention in the PCB studies and, given that one can never measure every possible confounder, the range of control variables that have been evaluated is impressive. We disagree with the authors' assertion that only standardized test scores are sufficiently reliable for use in these studies; behavioral teratogens often involve subtle effects, which can be identified most effectively by innovative, narrow‐band tests that have not yet been normed. Moreover, longitudinal statistical analysis is not necessarily the method of choice for the issues being addressed in this literature. One important new development that Cicchetti et al. fail to note is the emergence of evidence from both the Michigan and Dutch cohorts indicating that breast‐fed children are markedly less vulnerable. It is not yet clear to what degree this protective effect is attributable to nutrients in breast milk or to more optimal intellectual stimulation by nursing mothers, or both. However, the discovery of effect modifiers that can explain individual differences in vulnerability marks an important advance in our growing understanding of the teratogenic effects of exposure to environmental contaminants on child development. © 2004 Wiley Periodicals, Inc. Psychol Schs 41: 625–637, 2004.