Browsing by Subject "Structural Break"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • Thumbnail Image
    ItemOpen Access
    Joint test for structural model specification
    (2006) Yüksel, Serkan
    Aim of this thesis is to propose a test statistic that can test for true structural model in time series. Main concern of the thesis is to suggest a test statistic, which has joint null of unit root and no structural break (difference stationary model). When joint null hypothesis is rejected, source of deviation from the null model may be structural break or (and) stationarity. Sources of the deviation correspond to different structural models: Pure stationary model, trendbreak stationary model and trend-break with unit root model. The thesis suggests a test statistic that can discriminate null model from alternative models and more importantly, one alternative model from another. The test statistic that is proposed in the thesis is able to detect specific source of deviation from the null model. By doing so, we can determine the true structure model in time series. The thesis compares power properties of the test statistic that is proposed with the most favorable test in the literature. Simulation results indicate the power dominance over the test statistics in the literature. Moreover, we are able to specify true alternative model.
  • Thumbnail Image
    ItemOpen Access
    Performance analysis of single structural break test with an empirical study on efficient market hypothesis"
    (2005) Yıldız, İzzet
    In this thesis, performance of the single structural break tests is examined. Since it has proved superiority of Sequential F test on other single break tests, it is chosen as single break test. Monte Carlo simulation is run for different scenarios and performances of the test with respect to estimating break points, and parameters, and rejecting or accepting the joint null hypothesis is observed. For all cases small sample bias is observed. The test estimates parameters correctly for large samples but for small samples it underestimates or overestimates parameters. Another common problem is about joint null hypothesis. When test rejects the joint null, it doesn’t identify which of the joint hypothesis is rejected. Therefore in this study, we utilize the t-statistic of the parameters to determine the individual hypothesis rejected. In addition to these common problems we illustrate other scenario specific problems in this study. We examine the implications of our Monte Carlo findings by applying the break test to real life data and investigate the efficient market hypothesis using stock market data on SP&500. Application of the sequential F test shows evidence against the efficient market hypothesis.