依数据进行统计建模的三个基本分析层次

Three Basic Analysis Levels of Statistical Modeling based on Data

统计研究的原动力大致上可以归结为四个基本类型,即理论驱动、应用驱动、问题驱动和数据驱动。这些驱动力导致了统计研究所采用的范式各有不同,如理论推导、案例分析、数据模拟等,几乎每一种研究范式都包含了统计建模的内容和要求。本文提出统计建模的三个基本分析层次,即数据呈现、自由拟合与约束效应分析,并且尝试评述了各层级之间的递进关系,试图定义出一个从实践角度判断统计模型优劣的简单标准。

The motivation of statistical study generally comes down to four basic types：theorydriven,application-driven,problem-driven and data-driven.These driving forces lead to the research of statistics using different paradigms,such as theoretic deduction,case analysis,data modeling and so on,and almost every study of the statistical modeling paradigm contains the contents and requirements of statistical model.The paper puts forward statistical modeling of the three basic analysis levels：the data show,free fitting and restraint effectiveness analysis,and tries to comment a progressive relationship between the various hierarchy to try to define a simple standard to judge the merits of the statistical model from a practical point of view.