Mathematical Sciences: Statistical Graphics: Foundations of Regression Graphics

实时、计算量大、图形化的回归方法是本研究的重点。特别是，它将强调标准显示的理论统计基础和含义，例如基于包括广义线性模型在内的一大类周参数模型的基于残差的曲线族。这项研究的目标之一是开发一种纯图形分析，使用不源于残差的模型评估方法。这些发展将作为基于lisp的计算环境xlisp-stat的扩展提供。量化数据的图形显示是科学研究的中心工具。尽管这些在历史上仅限于可以在一张二维纸张上呈现的任何东西，但现在计算技术通常允许三维图形显示。如何有效地、技术上有效地科学地使用这项技术，需要认真思考和合理的处理。反过来，这应该会导致对标准显示器的更好理解，以及更有效的显示器的构建。

Real-time, computationally intensive, graphical methods for regression are the focus of this research. In particular, it will emphasize the theoretical statistical foundations and the implications of standard displays such as the family of plots based on residuals for a broad class of peri-parametric models including generalized linear models. One goal of this research is to develop a purely graphical analysis with model assessment methods that do not originate in residuals. These developments will be available as extensions of the LISP-based computing environment XLISP-STAT. Graphical displays of quantitative data are a central tool of scientific inquiry. Although these have historically been limited to whatever could be presented on a two-dimensional sheet of paper, computing technology now routinely allows three- dimensional graphical displays. How to make effective and technically valid scientific use of this technology needs careful thought and a logical treatment. In turn, this should result in both a better understanding of standard displays and the construction of still more effective ones.