Linear Models for Compositions
Linear Models for Compositions
复制标题
组合物的线性模型
DOI：
--
复制
发表时间：
2013
期刊：
影响因子：
--
通讯作者：
R. Tolosana
中科院分区：
文献类型：
--
作者：
K. G. Boogaart;R. Tolosana
在线性模型中，组合可以扮演因变量和自变量的角色。在这两种情况下，线性模型的参数再次由与数据相同的单纯形组成。大多数经典线性模型的方法在这些组合线性模型中都有相似之处。本章讨论了关于这个主题的几个问题。什么是组合线性模型？如何可视化组合的依赖，已经多变量，与进一步的外部协变量？如何用组合线性模型来建模和检验这种相关性？潜在的假设是什么？我们如何验证这些假设呢？结果的成分解释是什么？如何使用线性模型提供具有检验、置信区间和预测区域的统计证据？如何可视化模型结果和模型参数？如何比较组合线性模型以及如何找到最合适的模型？
Compositions can play the role of dependent and independent variables in linear models. In both cases, the parameters of the linear models are again compositions of the same simplex as the data. Most methods for classical linear models have a close analog in these compositional linear models. This chapter addresses several questions on this subject. What are compositional linear models? How to visualize the dependence of compositions, already multivariable, with further external covariables? How to model and check such dependence with compositional linear models? What are the underlying assumptions? How can we check these assumptions? What is the compositional interpretation of the results? How to use linear models to provide statistical evidence with tests, confidence intervals, and predictive regions? How to visualize model results and model parameters? How to compare compositional linear models and how to find the most appropriate one?