Needs of the data snooping bias tests for time series researchers

Purpose: It aims to discuss the importance of the methods, particularly for the researchers who works in comparing the studies of profitability of trading rules, mutual & hege funds performance, investment advisory services, and multiple competing model forecasting. Key literature/ theoretical perspectives: The importance of data snooping has been emphasized as a serious problem in finance literature. Literally, the meaning of the data snooping is hard to be defined. Crack (1999) described that "Data snooping (mistaking spurious statistical relationships for genuine ones) is an important and dangerous by-product of the financial analysis”. The Data Snooping Bias test is to investigate whether the model’s excess return against benchmark is resulted by superior predictive ability. Methodology: When we examine the large numbers of forecasting models (for example, technical trading rules) with a given data set, there is a high probability that one such model will yield superior performance by accident. White (2000) developed the first statistical testing model known as the Stationary Bootstrap Reality Check(RC). Hansen (2005) proposed a studentized version of the superior predictive ability (SPA) test. Extending those two models, Romano and Wolf (2005) introduced an RC-based stepwise test (Stepwise RC) and Hsu, Hsu and Kuan (2010) an SPA-based stepwise test.2 page(s