...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Environmental toxicology and chemistry >HANDLING NONNORMALITY AND VARIANCE HETEROGENEITY FOR QUANTITATIVE SUBLETHAL TOXICITY TESTS
【24h】

HANDLING NONNORMALITY AND VARIANCE HETEROGENEITY FOR QUANTITATIVE SUBLETHAL TOXICITY TESTS

机译：定量次要毒性测试的处理非常态和方差异质性

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

The advantages of using regression-based techniques to derive endpoints from environmental toxicity data are clear, and slowly, this superior analytical technique is gaining acceptance. As use of regression-based analysis becomes more widespread, some of the associated nuances and potential problems come into sharper focus. Looking at data sets that cover a broad spectrum of standard test species, we noticed that some model fits to data failed to meet two key assumptions-variance homogeneity and normality-that are necessary for correct statistical analysis via regression-based techniques. Failure to meet these assumptions often is caused by reduced variance at the concentrations showing severe adverse effects. Although commonly used with linear regression analysis, transformation of the response variable only is not appropriate when fitting data using nonlinear regression techniques. Through analysis of sample data sets, including Lemna minor, Eisenia andrei (terrestrial earthworm), and algae, we show that both the so-called Box-Cox transformation and use of the Poisson distribution can help to correct variance heterogeneity and nonnormality and so allow nonlinear regression analysis to be implemented. Both the Box-Cox transformation and the Poisson distribution can be readily implemented into existing protocols for statistical analysis. By correcting for nonnormality and variance heterogeneity, these two statistical tools can be used to encourage the transition to regression-based analysis and the depreciation of less-desirable and less-flexible analytical techniques, such as linear interpolation.
机译：使用基于回归的技术从环境毒性数据得出终点的优势是显而易见的，而且这种高级的分析技术正逐渐被人们接受。随着基于回归分析的使用变得越来越普遍，一些相关的细微差别和潜在问题变得更加突出。在查看涵盖广泛标准测试物种的数据集时，我们注意到某些模型适合的数据未能满足两个关键假设-方差同质性和正态性-这是通过基于回归的技术进行正确的统计分析所必需的。未能满足这些假设的原因通常是由于浓度变化幅度减小，显示出严重的不良影响。尽管通常与线性回归分析一起使用，但是仅当使用非线性回归技术拟合数据时，响应变量的转换才不合适。通过对样本数据集的分析，包括小Lemna，Eisenia andrei（陆地earth）和藻类，我们表明所谓的Box-Cox变换和泊松分布的使用都可以帮助校正方差异质性和非正态性，因此允许非线性回归分析得以实施。 Box-Cox变换和Poisson分布都可以很容易地实现到现有协议中进行统计分析。通过校正非正态性和方差异质性，可以使用这两个统计工具来鼓励向基于回归的分析过渡，以及鼓励不希望的和较不灵活的分析技术（例如线性插值）贬值。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号