Gad S C
Arch Toxicol Suppl. 1982;5:256-66. doi: 10.1007/978-3-642-68511-8_45.
One of the areas of toxicology in which a wide variation occurs in the statistical procedures used to analyze experimental data is behavioral toxicology. Due to either limitations in statistical training in the toxicologists or to a lack of understanding of the underlying biological mechanisms on the part of the statisticians, data is frequently analyzed by methodologies which either do not have optimal characteristics of sensitivity and power or for which the underlying assumptions as to the nature of the data are not valid. To establish a firm basis for an identification of the optimal and most appropriate forms of statistical analysis of behavioral toxicology data (and to design efficient and sensitive studies), the four general types of data (observational scores, response rates, error rates, and times-to-endpoints) and one special class of data (teratology and reproduction) are examined in detail. The present practices as to statistical analysis are then reviewed and suggestions as to optimal methods (based on experience, with data sets presented as examples) are developed and presented. The underlying key to this entire process is to establish the biological and statistical nature of the data being generated and to design and analyze experiments accordingly.
行为毒理学是毒理学的一个领域，在用于分析实验数据的统计程序方面存在很大差异。由于毒理学家在统计培训方面存在局限性，或者统计学家对潜在的生物学机制缺乏了解，数据经常采用灵敏度和功效并非最佳，或者关于数据性质的基本假设无效的方法进行分析。为了为确定行为毒理学数据的最佳和最合适的统计分析形式建立坚实的基础（并设计高效且灵敏的研究），详细研究了四种一般类型的数据（观察分数、反应率、错误率和终点时间）以及一类特殊的数据（致畸学和生殖学）。然后回顾了当前的统计分析实践，并提出了关于最佳方法的建议（基于经验，并给出数据集作为示例）。整个过程的关键在于确定所生成数据的生物学和统计学性质，并据此设计和分析实验。
