Exploring Program of Factor and Cluster Analysis Based on Econometrics Panel Data and Its Application

本文对不能区分或者难以确定因变量和自变量的平板数据进行数据挖掘，综合使用因子分析和聚类分析方法来分类，首先我们假设在不同时间点上所有的变量具有相同的结构，这样我们可以运用因子分析来估计这个结构，通过因子分析得到因子得分，然后计算每个样本得分，根据得分计算每个个体的均值、标准差、伪信噪比（PSTN）和趋势差（TD），由4个指标对个体进行分类。因为平板数据具有时间特性，我们利用时间序列进行预测，从而计算出残差，根据残差和上面4个指标来对个体进行分类。同时，我们提出一种新的标准来确定因子个数以配合因子分析在面板数据使用，来更好的充分的利用信息。利用事例来验证我们的思路和分析流程，结果显示出了具有一定优越性。

關鍵字

因子分析；聚类分析；平板数据；发展

並列摘要

This paper synthesizes factor analysis and cluster analysis method on the panel data samples which it difficult to ensure and distinguish the independent variable from dependent variable. Factor analysis method analyses the data to get the common structure of samples in each time-point which we can assume the cases of panel data have, then we figure out score of each sample at each time-point. Finally, according to the scores, we calculate the mean, standard deviation, pseudo-signal-to-noise (PSTN) and tendency deviation (TD) of each sample to do cluster analysis under different situations. Furthermore, we utilize the time-series models to predict, we calculate a statistics on the residual, we accord to mean, standard deviation, PSTN, TD and the new statistics to classify the samples. Meanwhile, we develop a criterion how to determine the size of factor to be suitable for doing factor analysis in panel data to utilize more information. A case is used to verify the analysis program.