基於逐步移除設限法可以允許試驗者收集到較長產品的壽命的優點，以及區間設限法具有容易操作之特性，本論文針對逐步型一區間設限資料，進行模型參數最大參數估計解根方法的穩定性評估。假設產品的壽命服從雙參數的Burr XII分配，分別使用擬牛頓法及啟發式演算法中的基因演算法 (GA) 與差分演算法 (DE) 來求解模型參數的最大概似估計量。藉由統計模擬研究計算參數估計值的偏誤及均方誤，以評估三種不同解根方法的穩定度。研究中發現，GA 及DE在最大參數估計量的解根表現中有較穩定的運算結果。 To the advantage of progressively censoring method that allow experimenters to observe more extreme lifetimes of products and the administration convenience of the interval censoring method, this thesis evaluates the performance of difference algorithms, the quasi-Newton method, genetic algorithm method and differential evolution methods for maximum likelihood estimates based on the progressively type I interval censored samples for the Burr type XII distribution. An intensive simulation study was conduct for comparing the estimation performance of the quasi-Newton method, genetic algorithm (GA) method and differential evolution (DE) methods. We found that the GA method and DE methods outperform the quasi-Newton method for most cases in the simulation study.
