本篇文章之主要目的是將Efron於1979年所提出之拔靴法(Bootstrap method)應用在非常態ARMA(p,q)模式上。我們考慮的結構包括AR(1)，AR(2)，MA(1)，MA(2)，ARMA(1,1)，而考慮的非常態干擾分配則包括對數常態分配，均勻分配，迦瑪分配以及指數分配，對每一個模式，所設定之參數值組合涵蓋了使模式平穩及/或可逆之參數組合中的重要區域。我們比較傳統的最小平方法與利用500個拔靴重複子(Bootstrap repetitions)的拔靴法在參數估計上的差異。進一步我們也以MAE及MAPE等準則比較了這兩種估計模式在預測上的優劣。模擬結果顯示，在參數估計方面，當所設定的參數範圍接近非平穩或非可逆條件時，拔靴法所獲得的參數估計值表現會較佳。然而在預測效益方面，利用往前一期預測法，並配合拔靴法重新改造的數列，在具有非常態干擾項AR(1)，AR(2)以及ARMA(1,1)模式的預測效益上的確有較佳的效果。 關鍵詞：拔靴法，非常態ARMA(p,q) ，最小平方法 / In this thesis, the Bootstrap technique proposed by Efron in 1979 is applied to parameter estimation and forecasting of non-normal ARMA(p,q) time series models. A simulation study is conducted where artificial time series are generated from various ARMA structures with different non-normal noise distributions. The ARMA structures considered in the simulation are AR(1), AR(2), MA(1), MA(2) and ARMA(1,1), while the non-normal noise distributions include Log-normal, Uniform, Gamma, and Exponential distributions. For each structure, the parameter values used cover important regions of the stationary and/or invertible parameter space . The conventional least-square estimators of the parameters are compared with the corresponding non-parametric Bootstrap estimator, obtained by using 500 Bootstrap repetitions for each series. Furthermore, forecasts based on these estimated model are also compared by using such criteria as MAE and MAPE . KEY WORDS bootstrap; non-normal ARMA(p,q); least-square

Most recorded data are rounded to the nearest decimal place due to the precision of the recording mechanism. This rounding entails errors in estimation and measurement. In this paper, we compare the performances of three types of estimators based on rounded data from time series models, namely A-K corrected estimator, approximate MLE and the SOS estimator. In order to perform the comparison, the A-K corrected estimators for the MA(1) model are derived theoretically. To improve the efficiency of the estimation, two types of variance-reduction estimators are further proposed, which are based on linear combinations of aforementioned three estimators. Simulation results show the proposed variance reduction estimators significantly improve the estimation efficiency.

Nesse trabalho, exploramos técnicas para análise de séries temporais para fins de previsão. Para tanto, foram considerados dados observados de três séries climáticas e de uma série econômica. Para análise das séries climáticas, foi considerada a modelagem multivariada em comparação com os subsequentes modelos univariados de cada série. Os modelos multivariados e univariados foram comparados com base em seus respectivos resultados preditivos. Para análise da série econômica, considerou-se a modelagem ARMA-GARCH, cuja média condicional e variância condicional são modeladas conjuntamente. Para essa mesma série foi realizada uma modelagem ARIMA em que considerou-se dois casos. No primeiro, a modelagem foi realizada na série original. No segundo, foi realizada na pré-modelagem uma filtragem na série, denominada de sistema de decomposição Wavelet- WavDS, com o objetivo de melhorar o poder preditivo. Na seleção dos modelos ARIMA, considerou-se a metodologia backtesting, em que as previsões são realizadas de forma sequencial, o modelo selecionado foi o que apresentou menor raiz quadrada do erro quadrático médio de previsão (REQM). Toda análise estatística realizada nesse trabalho foi com auxílio do software livre R. / In this study, we explored techniques of time-series analysis for prediction purposes. For that, we considered data observed from three climate series and one economic series. For the analysis of the climate series, we considered the multivariate modelling in comparison with the subsequent univariate models of each series. The multivariate and univariate models were compared based on their respective predictive results. For the analysis of the economic series, the ARMA-GARCH modeling was considered, whose conditional average and conditional variance are modeled together. For this same series, the ARIMA modeling was used, considering two cases. At first, the modeling was performed in the original series. In the second, we carried out a filtering in the series during pre-modeling, called Wavelet- WavDS decomposition system, in order to improve the predictive power. In the selection of ARIMA models, we considered the backtesting methodology in which forecasts are performed in sequence. The model selected showed the lowest square root mean of the prediction square error (REQM). All statistical analyses performed in this work were carried out using the free software R.

This thesis has used the Fama and French five-factor model (FF5M) and proposed an alternative model. The proposed model is named the Fama and French five-factor heteroscedastic student's model (FF5HSM). The model utilises an ARMA model for the returns with the FF5M factors incorporated and a GARCH(1,1) model for the volatility. The FF5HSM uses returns data from the FF5M's portfolio construction for the Japanese stock market and the five risk factors. The portfolio's capture different levels of market capitalisation, and the factors capture market risk. The ARMA modelling is used to address the autocorrelation present in the data. To deal with the heteroscedasticity in daily returns of stocks, a GARCH(1,1) model has been used. The order of the GARCH-model has been concluded to be reasonable in academic literature for this type of data. Another finding in earlier research is that asset returns do not follow the assumption of normality that a regular regression model assumes. Therefore, the skewed student's t-distribution has been assumed for the error terms. The result of the data indicates that the FF5HSM has a better in-sample fit than the FF5M. The FF5HSM addresses heteroscedasticity and autocorrelation in the data and minimises them depending on the portfolio. Regardingforecasting, both the FF5HSM and the FF5M are accurate models depending on what portfolio the model is applied on.

本研究有系統地收集了2003年8月12日到2005年9月30日止共495個交易日的台指期貨、選擇權市場裡P/C量、P/C倉、隱含波動率(AIV)、不同天數的歷史波動率等收盤資料，進行這些因素與行情走勢間的關係，以及因素彼此的互動性。結果證實分析台指選擇權指標是需要區分金融重大衝擊前後期間，以及區分漲勢、跌勢、盤整的各期間，各期間的選擇權指標均會有不同意涵。 本論文證實使用結構轉換的Chow-ARMA(2,1)模型可能比較符合模擬指數 實況，且GARCH(1,1) 模型也很適合描述台期指貨波動度預測力。在選擇權指標方面：P/C量與AIV與台指期貨呈現負相關，P/C倉與台指期貨正相關。其中以P/C倉對指數漲跌的影響程度最大、P/C量的影響程度次之、AIV影響程度最小。若把隱含波動率區分成買權與賣權之各個波動率更有效地預測行情走勢，在大跌期間的買賣權隱含波動率更能表現出優越的預測能力，其中前兩期的賣權隱含波動率(PIV)更是效率性指標， 實證結果使用20天的歷史波動率比較能貼近選擇權市場的變化，跟過去教 科書慣用的90天不同。若比較歷史波動率與隱含波動率間的關係，結論是當「大跌期」歷史波動率大於買權隱含波動率(CIV)時，買權是會被低估的，其他的各種假設條件均不成立。理由有二：一是市場效率性決定了是否可使用隱含波動率與歷史波動率之間的高低關係。二是「大跌時期」相對於「大漲時期」的市場資訊被反應的更敏銳，而在「大跌時期」的賣權價格反應比買權價格反應更快速敏銳。 本研究推論的Chow-ARMA(2,1) 台指期貨模型、GARCH(1,1) 波動率模型、P/C量-P/C倉-AIV的多變數模型、FMA20/XIV模型等等在研判指數變化上具有參考價值，進一步均可以做為選擇權操作策略參考依據之一。