skip to main content

Title: Statistical analysis of litter experiments in teratology

Teratological data is binary response data (each fetus is either affected or not) in which the responses within a litter are usually not independent. As a result, the litter should be taken as the experimental unit. For each litter, its size, n, and the number of fetuses, x, possessing the effect of interest are recorded. The ratio p = x/n is then the basic data generated by the experiment. There are currently three general approaches to the analysis of teratological data: nonparametric, transformation followed by t-test or ANOVA, and parametric. The first two are currently in wide use by practitioners while the third is relatively new to the field. These first two also appear to possess comparable power levels while maintaining the nominal level of significance. When transformations are employed, care must be exercised to check that the transformed data has the required properties. Since the data is often highly asymmetric, there may be no transformation which renders the data nearly normal. The parametric procedures, including the beta-binomial model, offer the possibility of increased power.
Authors:
;
Publication Date:
OSTI Identifier:
6702512
Report Number(s):
PNL-4425
ON: DE83002764
DOE Contract Number:
AC06-76RL01830
Resource Type:
Technical Report
Research Org:
Pacific Northwest Lab., Richland, WA (USA)
Country of Publication:
United States
Language:
English
Subject:
63 RADIATION, THERMAL, AND OTHER ENVIRON. POLLUTANT EFFECTS ON LIVING ORGS. AND BIOL. MAT.; TERATOGENESIS; STATISTICS; DATA PROCESSING; LABORATORY ANIMALS; MATHEMATICS; PROCESSING 560305* -- Chemicals Metabolism & Toxicology-- Vertebrates-- (-1987)