本三年期專題研究包含三個方向分別為：（一）概似多維報酬函數 (Approximation of multivariate return distribution)；（二）資產定價模 型測試(CAPM tests)；（三）股市中陰曆效應(Lunar calendar effects in East Asian equity markets)。在第一個研究方向中，嘗試掌握捕捉多維投資組合報 酬率分佈函數方面新進的突破方法，或是將單維方面之努力推展至多維層次， 並嘗試推廣應用，議題至少包括投資組合風險值與量化營運風險等。在第二個 研究方向中，則是就Sharpe 與Lintner 所提之資產定價模型運用份量迴歸與Bai 與 Perron 所提多重結構性變化架構，進行檢定測試。最後，有別於傳統以西洋日曆為 主所做之以日曆效應研究，轉以農民曆為主來探討東亞股市，是否呈現不同型態之 市場異常，為求慎重並且進行相關新進檢定，以避免出現資料蒐評之困境。 This three-year term proposal includes three major research directions: (1) Approximation of multivariate return distribution; (2) CAPM tests; (3) Lunar calendar effects in East Asian equity markets. The first direction overviews the most updated approaches in capturing the multivariate return distribution, e.g. multivariate extreme value theory, maximum entropy or minimum cross-entropy econometrics, and Wang』s transformation. Further, it does not only extend the innovative approaches in univariate case but also compare their performance. Its various applications include at least portfolio value-at-risk, hedging, and quantitative operational risk. The second direction employs quantile regression technque and Bai and Perron』s multiple structural change framework to conduct CAPM tests under Sharpe-Linter version. Different from previous studies on solar calendar system, the third direction focuses on lunar calendar effects in East Asian equity markets. X-12-ARIMA and ordered probit models are employed and the analysis outcome and its cultural implications are interpreted. Some tests are applied to prevent data snooping biases, e.g. White』s reality check and Hansen』s SPA test.
