藉由K 均值分群與階層式分群程序對潛在群體分析做參數估計

Abstract

本研究的主要目的是藉由群聚分析的方法對潛在群體模型做參數估計。我們引用了群聚方法中的k 均值分群和階層式分群的想法，將原本的距離測度改成相關係數或共變異數，然後對所有的主體分群，使得屬於在同一群的主體所測得的各項目能互相獨立。將估計出的潛在群體視為已知變數後，再去估計潛在群體迴歸分析模型的參數就變得容易多了。我們的模擬結果顯示出：所用的測度為相關係數或共變異數的k 均值分群法表現得不錯，但是所用的測度為共變異數的階層式分群法表現得並不好。

The aim of the study is to estimate the parameters of the latent class models via clustering methods. We use k-means and hierarchical ideas of clustering methods with the correlation (or covariance) among items as the distance measure to group objects such that, for all objects who belong to the same latent class, items are ”independent”. By viewing the estimated latent class as known variable, it becomes easy to estimate the parameters in the regression extension of latent class analysis (RLCA) model. The results of our simulation study display that the k-means method with the correlation (or covariance) measurement performed well, but the hierarchical method with the covariance measurement didn’t perform well.