指數迴歸分析之參數估計

THE ESTIMATION OF PARAMETERS OF EXPONENTIAL REGRESSION ANALYSIS

指數多元迴歸分析中，其參數的估計往往利用最大概似估計量（maximum likelihood estimator），或者採用Cox的偏概似估計量（cox partial likelihood estimator）來估算，通常均是較為繁雜；本文運用次序統計量提供一個近似最大概似估計量，其不必用重覆計算程序，而其有效性也與最大概似估計量沒有顯著的差異，其實務上應較其他方法簡便有效。

By maximum likelihood estimator (MLE), and/or Cox's partial likelihood estimator are regularly employed to estimate parameters in multiple exponential regression analysis. But their processes are generally considered more complicated. In this study, we employed order statistics and provided an asymptotic maximum likelihood estimator (AMLE). The AMLE does not require an iterative calculating process. So, the efficiency of AMLE is not significantly different from MLE, and the method is more convenient and efficient than other methods in application.