Averaging Models: Parameters Estimation with the R-Average Procedure.
Averaging Models: Parameters Estimation with the R-Average Procedure.
复制标题
平均模型：使用 R 平均过程进行参数估计。
作者：
G. Vidotto;D. Massidda;S. Noventa
在信息整合理论（安德森，1981，1982）的理论框架内提出的功能测量方法，可以是一个有用的多属性分析工具。与大多数统计模型相比，平均模型可以在不增加复杂性的情况下考虑相互作用效应。可以使用RADIOR方法（Vidotto & Vicentini，2007）来估计这些模型的参数。通过在模型选择过程中使用多个信息标准，RADIOR允许识别说明数据的参数的最佳子集。在回顾了一般方法之后，我们提出了一个在Rproject框架下的程序实现，然后使用Monte Carlo方法进行了B y一些实验。多属性模型一般遵循三个步骤：属性的评估，所获得的主观维度的整合，和一个结论性的阶段。在最后一个阶段，前面的过程的结果被转化为一个排名顺序，一组成对的偏好或在某个真实的区间上的评级（Lynch，1985; Oral & Kettani，1989）。安德森（Anderson，1981，1982）提出了这些模型的一个子集，将平均过程确定为广泛使用的认知整合规则之一。平均过程使用权重和比例值参数表示。比率量表用于衡量体重，等间隔量表用于衡量价值（Zalinski &安德森，1989）。此外，Norman（1976）和安德森（1982）提出的子设计方法，通过将选定的子设计与完全事实设计相连，使这些参数完全可识别。例如，完全三向设计（A × B × C）可补充艾德三向
The Functional Measurement approach, proposed within the theoretical framework of Information Integration Theory (Anderson, 1981, 1982), can be a useful multiattribute analysis tool. Compared to the majority of statistical models, the averaging model can account for interaction effects without adding complexity. The RAverage method (Vidotto & Vicentini, 2007) can be used to estimate the parameters of these models. By the use of multiple information criteria in the model selection procedure, RAverage allows for the identification of the best subset of parameters that account for the data. After a review of the general method, we present an implementation of the procedure in the framework of Rproject, followed b y some experiments using a Monte Carlo method. Multiattribute models generally follow three steps : evaluation of the attributes, integration of the obtained subjective dimensions, and a conclusive stage. In the last stage, the results of the previous processes are transformed into a ranking order, a set of pairwise preferences or a rating over some real interval (Lynch, 1985; Oral & Kettani, 1989). A subset of these models, proposed by Anderson (1981, 1982), identifies the averaging process as one of the widely used cognitive integration rules. The averaging process uses a weight and scale value parameters representation. Ratio scales are involved in the measurement of weights and equalinterval scales are used to measure values (Zalinski & Anderson, 1989). Furthermore, the method of subdesigns, proposed by Norman (1976) and Anderson (1982), allows for complete identifiability of these parameters by adjoining selected subdesigns to the full facto rial design. For instance, a full threeway design (A × B × C) can be supplement ed with three twoway