1 |
一種基於函數型資料主成分分析的曲線對齊方式 / A Curve Alignment Method Based on Functional PCA林昱航, Lin,Yu-Hang Unknown Date (has links)
函數型資料分析的是一組曲線資料,通常定義域為一段時間範圍。常見的如某一個地區人口在成長期的身高紀錄表或是氣候統計資料。函數型資料主要特色曲線間常有共同趨勢,而且個別曲線反應共同趨勢時也有時間和強度上的差異。本文研究主要是使用Kneip 和 Ramsay提出,結合對齊程序和主成分分析的想法作為模型架構,來分析函數型資料的特性。首先在對齊過程中,使用時間轉換函數(warping function),解決觀測資料上時間的差異;並使用主成分分析方法,幫助研究者探討資料的主要特性。基於函數型資料被預期的共同趨勢性,我們可以利用此一特色作為各種類型資料分類上的依據。此外本研究會對幾種選取主成分個數的方法,進行綜合討論與比較。 / In this thesis, a procedure combining curve alignment and functional principal component analysis is studied. The procedure is proposed by Kneip and Ramsay .In functional principal component analysis, if the data curves are roughly linear combinations of k basis curves, then the data curves are expected to be explained well by principle component curves. The goal of this study is to examine whether this property still holds when curves need to be aligned. It is found that, if the aligned data curves can be approximated well by k basis curves, then applying Kneip and Ramsay's procedure to the unaligned curves gives k principal components that can explain the aligned curves well. Several approaches for selecting the number of principal components are proposed and compared.
|
2 |
貝氏曲線同步化與分類 / Bayesian Curve Registration and Classification李柏宏, Lee,Po- Hung Unknown Date (has links)
函數型資料分析為近年發展的統計方法。函數型資料是在一段特定時間上,我們只在離散的時間點上收集觀測值。例如:氣象觀測站所收集到的每月氣溫、雨量資料,即是一種常見的函數型資料。函數型資料主要有三種特色,共同趨勢性、觀測個體反應強度不同,觀測個體時間特色上的差異。本文研究主要是使用,Brumback與Lindstrom在2004所提出的自模型迴歸族(self-modeling)當作模型架構來處理函數型資料的趨勢性與個體反應強度。而為了處理函數型資料的時間差異性,我們在模型中加入時間轉換函數(time transformation function),處理函數型資料的時間差異性步驟,這個過程稱為同步化。經過同步化的處理後,能幫助研究者更清楚資料的特性。模型中除了時間轉換函數的部份,其餘模型中的參數我們是利用馬可夫鏈蒙地卡羅法中的Gibbs Sampling來進行參數的抽樣,並以取出的抽樣值來估計參數。時間轉換函數的部份,我們使用概似懲罰函數(penalized likelihood function)來估計時間轉換函數的參數部份。由於函數型資料擁有趨勢性,我們預期不同類別的資料,會呈現不同的趨勢性,我們將利用此一特色當做分類上的標準。
關鍵詞:函數型資料分析、曲線同步化、曲線區別分析、馬可夫鏈蒙地卡羅法。 / Functional data are random curves observed in a period of time at discrete time points.They often exhibit a common shape, but with variations in amplitude and phase across curves.To estimate the common shape,some adjustment for synchronization is often made,which is also known as time warping or curve registration.In this thesis,splines are used to model the warping functions and the common shape. Certain parameters are allowed to be random.For the estimation of the random parameters,priors are proposed so that samples from the posteriors can be obtained using Markov chain Monte Carlo methods.For the estimation of non-random parameters, a penalized likelihood approach is used. It is found via simulation studies that for a set of random curves with a common shape,the estimated common shape function looks like the true function up to a location-scale transform,and the curve alignment based on estimated time warping functions looks reasonable.For two groups of random curves which differ in the group common shape functions,synchronization also improves the discrimination between groups in some cases.
Key words: functional data analysis,curve registration,curve discrimination,markov chain monte carlo method.
|
Page generated in 0.0217 seconds