本計畫將提出幾個函數型指標來決定不規則抽樣下的長期追蹤資料(longitudinal data) 之最適分群數。當資料型態為長期在不同時間點所紀錄的觀測值時,可以()代表觀測資料,其中為第i個觀測對象在第ijijyt,ijyj個時間點所得到的變數測量值,ijtimj,...,1=,ni,...,1=。我們假設為隨機過程在間斷時間點所得到的觀測值,而且此個的觀測對象具有潛在的分群結構。根據此假設,本計畫即以函數型資料的概念來修正傳統多變量方法中的群數選取指標。透過適當的無母數平滑估計方法,我們所提出的函數型指標將可應用在不規則抽樣或稀疏抽樣下的長期追蹤資料。本計劃將透過實際資料與模擬驗證的方式討論各函數型指標的可行性與有效性,並進一步做理論探討與發展其他新指標。 We propose several functional indices for determining the number of clusters in irregularly sampled longitudinal data set. The data are viewed as realizations of a mixture of stochastic processes and each sub-process corresponds to a cluster. The functional indices are modified from the objective functions of traditional multivariate stopping rules. Through nonparametric smoothing techniques, our proposed method can be applied into the irregularly sampled data or spare data. The proposed method will be evaluated by the simulation study and data examples. Furthermore, we will try to investigate the theoretical properties and develop other new indices.