The putative effects of smoking by pregnant women on birthweight, IQ, and developmental disabilities in their infants : a methodological review and multivariate analysis /
Concern over smoking by pregnant women has increased in the last decade. Concomitantly, research addressing the possible effects of smoking, and of smoking while pregnant in particular, has grown. Research reports attribute to smoking numerous harmful effects, including reduced weight at birth and l...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
2000.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=727853271&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | Concern over smoking by pregnant women has increased in the last decade. Concomitantly, research addressing the possible effects of smoking, and of smoking while pregnant in particular, has grown. Research reports attribute to smoking numerous harmful effects, including reduced weight at birth and later psychological problems. Some researchers, however, dispute such interpretations. The present report begins with a historical and critical review of criteria for establishing causality and a discussion of research designs that permit causal inference. Next, multivariate analyses assess the methodological quality of existing studies. The report also includes a methodological review and a reanalysis of Li et al. (1993). Gosset's t, Mann-Whitney U, and MANOVA results indicated a small, statistically nonsignificant association between causal purpose of study and design appropriate for causal inference. Thus, researchers identified the nature or purpose of a study as causal whether or not the design of the study was appropriate to infer causation. Similarly, results of the median test, Kruskal-Wallis ANOVA, and Kendall's tau-b, with Welch's t and ANCOVA results, were small and nonsignificant for design and causal conclusion. Thus, researchers drew causal inferences from studies not designed for such inferences as often as from suitably designed studies. The median test and Kruskal-Wallis one-way ANOVA, with Kendall's tau-b and the Mann-Whitney U, indicated a moderate, statistically significant association between causal assertions and causal conclusion. Partial correlations behaved similarly. Thus, authors who indicated that maternal smoking was already found to be causative made causal inferences from their own work. Finally, partial correlations, with Mann-Whitney and Kruskal-Wallis results, were approximately moderate and statistically significant for causal assertions with causal conclusion. These results remained unchanged with design as an independent variable and as a control. Thus, causal interpretations of research were associated with a prior assumption of causation, and the appropriateness of a study's design for such interpretations made little difference for this association. A suggested conclusion is that causal inferences depend on a prior causal assumption, not on an appropriate design. In addition, the results provide little evidence for an effect of smoking on IQ, birth weight, or the developmental disabilities examined. |
|---|---|
| Item Description: | Vita. "Major Subject: Educational Psychology". |
| Physical Description: | xii, 154 leaves ; 28 cm. Issued also on microfiche from University Microfilm Inc. |
| Bibliography: | Includes bibliographical references (leaves 129-148). |