2002 / October

潛在變項選擇模型結構方程模型之最大概似估計
Maximum Likelihood Estimation of Structural Equation Models with Latent Variable Selection Model

研究論文 Article

鄭中平, 翁儷禎
Chung-Ping Cheng, Li-Jen Weng

5-27

本研究目的在以MCEM 算則（Monte Carlo expectationmaximization Algorithm）進行潛在變項影響資料遺漏時結構方程模型的參數估計。結構方程模型分析常需實徵資料驗證研究者的假設模型，資料發生遺漏是其收集過程經常遭遇的情形。遺漏可能與外顯變項有關，但亦可能與潛在變項有關。Muthén、Kaplan與Hollis（1987）描述了外顯變項或潛在變項影響遺漏與否的遺漏機制模型，並發現多數情形下為不可忽略遺漏，現行之遺漏值處理法未必適用。因此，本研究針對Muthén等人之遺漏機制模型發展結構方程模型參數估計方法，並以實例比較其與常用遺漏值處理法的差異，初步發現本研究建議方法在潛在變項影響資料遺漏情形下表現最佳。

The Monte Carlo expectation-maximization algorithm was proposed for parameter estimation with latent variable selection model in structural equation modeling. Latent variables are allowed to influence data missingness in latent variable selection model. This missing mechanism in most cases is not missing at random. The missing data treatment methods available at present therefore may not be applicable and new development is called for. An empirical example of latent variable selection model was presented. The results indicated that the proposed method yielded satisfactory parameter estimates.

非隨機遺漏, 選擇模型, 結構方程模型, 最大概似法, 潛在變項
nonignorable missingness, selection model, structural equation modeling, maximum likelihood estimation, latent variable

心理學
Psychology

https://doi.org/10.7014/SR.2002100001

Full-text