Compensating for Deficiencies in Perinatal Data Sets: Parametric Perspectives (From Treatment for Drug-Exposed Women and Their Children: Advances in Research Methodology, P 272-291, 1996, Elizabeth R Rahdert, ed. -- See NCJ-163710)

• NCJ Number: 163727
• Author(s): A S Tippetts; P R Marques
• Date Published: 1996
• Length: 20 pages
• Sale Source: National Institute on Drug Abuse; NCJRS Photocopy Services
• Publication Type: Report (Technical)
• Language: English
• Country: United States of America
• Note: DCC. NIDA Research Monograph Series 165

Annotation

New techniques are described that may be useful in dealing with difficult or problematic data sets on perinatal drug use and that attempt to resolve data deficiencies and irregularities.

Abstract

Before any hypothesis testing is undertaken in research studies, data should always be examined to assess the degree to which assumptions may be violated, especially assumptions requiring a data set to be normally distributed. Even when a data set includes many variables, the time involved in quickly checking the distribution of each variable can pay big dividends in the long run. Histograms represent the most basic visual check, and many statistical software packages can superimpose the outline of the normal distribution over the histogram. Abnormal distributions should be excluded from parametric analysis altogether unless they can be modified in some way to become moderately normal. Parametric perspectives on ways of compensating for deficiencies in perinatal data sets are examined that relate to conventional data transformations, maximizing normality, biserial change variables, data errors, missing data, and nonlinear relationships. The authors note that the most common data deficiencies result from the researcher's imperfect methods of scaling, measuring, and collecting data and from the use of overly simplistic perspectives in modeling and diagnosing data relationships. 9 references and 3 figures