This thesis consists of five self-contained papers, which are all related to the financial markets in the three Baltic States, Estonia, Latvia and Lithuania.  Paper [I] studies the impact of news from the Moscow and New York stock exchanges on the returns and volatilities of the Baltic States' stock market indices using a time series model that accounts for asymmetries in the conditional mean and variance functions. We find that news from New York has stronger e¤ects on returns in Tallinn. High-risk shocks in New York have a stronger impact on volatility in Tallinn, whereas volatility in Vilnius is more in.uenced by high-risk shocks from Moscow. Riga does not seem to be affected by news arriving from abroad. Paper [II] suggests a nonlinear and multivariate time series model framework that enables the study of simultaneity in returns and in volatilities, as well as asymmetric effects arising from shocks and exogenous variables. The model is employed to study the three Baltic States' stock exchanges. Using daily data, we find recursive structures, with returns in Riga, directly depending on returns in Tallinn and Vilnius, and Tallinn on Vilnius. For volatilities, both Riga and Vilnius depend on Tallinn. Paper [III] studies the link between political news, and the returns and volatilities in the Baltic States' stock markets. We find that domestic and foreign non-Russian political news led, on average, to lower uncertainty in the stock markets of Riga and Tallinn in 2001-2003. At the same time, political risk from Russia increased the volatility of the stock market in Tallinn. There is a weak relationship between political risk and the stock market volatility in the Baltic countries in 2004-2007. Paper [IV] studies the impact of market jumps on the time varying return correlations between stock market indices in the Baltic countries. An EARJI-EGARCH model facilitating direct modeling of the time varying return correlations is introduced. The empirical results indicate that there are quite a large number of identified jumps in the emerging Baltic States' stock markets. Isolated market jumps in one of the markets generally have no or small e¤ects on the time-varying correlations. In contrast, simultaneous jumps of equal sign increase the average correlation, in some cases by as much as 100 percent. In Paper [V] the hypothesis that financial development promotes economic growth is tested for the three Baltic countries using a time series approach that allows for interactions between the countries. We find that economic growth is a positive function of financial development, proxied by the amount of bank credit to the private sector, in the long run. The results also show that there is long run interaction between the three Baltic countries.

The aim of the study was to see if any relationship between government spending andunemployment could be empirically found. To test if government spending affectsunemployment, a statistical model was applied on data from Sweden. The data was quarterlydata from the year 1994 until 2012, unit-root test were conducted and the variables wheretransformed to its first-difference so ensure stationarity. This transformation changed thevariables to growth rates. This meant that the interpretation deviated a little from the originalgoal. Other studies reviewed indicate that when government spending increases and/or taxesdecreases output increases. Studies show that unemployment decreases when governmentspending/GDP ratio increases. Some studies also indicated that with an already largegovernment sector increasing the spending it could have negative effect on output. The modelwas a VAR-model with unemployment, output, interest rate, taxes and government spending.Also included in the model were a linear and three quarterly dummies. The model used 7lags. The result was not statistically significant for most lags but indicated that as governmentspending growth rate increases holding everything else constant unemployment growth rateincreases. The result for taxes was even less statistically significant and indicates norelationship with unemployment. Post-estimation test indicates that there were problems withnon-normality in the model. So the results should be interpreted with some scepticism.

The field of time-series analysis has made important contributions to a wide spectrum of applications such as tide-level studies in hydrology, natural resource prospecting in geo-statistics, speech recognition, weather forecasting, financial trading, and economic forecasts and analysis. Nevertheless, the analysis of the non-Gaussian and non-stationary features of time-series remains challenging for the current state-of-art models. This thesis proposes an innovative framework that leverages the theory of copula, combined with a probabilistic framework from the machine learning community, to produce a versatile tool for multiple time-series analysis. I coined this new model Kernel-based Copula Processes (KCPs). Under the new proposed framework, various idiosyncracies can be modeled compactly via a kernel function for each individual time-series, and long-range dependency can be captured by a copula function. The copula function separates the marginal behavior and serial dependency structures, thus allowing them to be modeled separately and with much greater flexibility. Moreover, the codependent structure of a large number of time-series with potentially vastly different characteristics can be captured in a compact and elegant fashion through the notion of a binding copula. This feature allows a highly heterogeneous model to be built, breaking free from the homogeneous limitation of most conventional models. The KCPs have demonstrated superior predictive power when used to forecast a multitude of data sets from meteorological and financial areas. Finally, the versatility of the KCP model is exemplified when it was successfully applied to non-trivial classification problems unaltered.

University of Macau / Faculty of Science and Technology / Department of Computer and Information Science

University of Macau / Faculty of Science and Technology / Department of Mathematics

University of Macau / Faculty of Science and Technology / Department of Mathematics

The field of time-series analysis has made important contributions to a wide spectrum of applications such as tide-level studies in hydrology, natural resource prospecting in geo-statistics, speech recognition, weather forecasting, financial trading, and economic forecasts and analysis. Nevertheless, the analysis of the non-Gaussian and non-stationary features of time-series remains challenging for the current state-of-art models. This thesis proposes an innovative framework that leverages the theory of copula, combined with a probabilistic framework from the machine learning community, to produce a versatile tool for multiple time-series analysis. I coined this new model Kernel-based Copula Processes (KCPs). Under the new proposed framework, various idiosyncracies can be modeled compactly via a kernel function for each individual time-series, and long-range dependency can be captured by a copula function. The copula function separates the marginal behavior and serial dependency structures, thus allowing them to be modeled separately and with much greater flexibility. Moreover, the codependent structure of a large number of time-series with potentially vastly different characteristics can be captured in a compact and elegant fashion through the notion of a binding copula. This feature allows a highly heterogeneous model to be built, breaking free from the homogeneous limitation of most conventional models. The KCPs have demonstrated superior predictive power when used to forecast a multitude of data sets from meteorological and financial areas. Finally, the versatility of the KCP model is exemplified when it was successfully applied to non-trivial classification problems unaltered.

There is a lack of measured, long-term, reliable, and well-distributed hydrological variables in Canada. These hydrological variables include, but are not limited to: temperature, precipitation, ground runoff, evapotranspiration, soil moisture, and snow water equivalent. The objective of this thesis was to establish the best possible distributed estimates of these hydrological variables for Canada over the period of 1961-2000. The first step was to interpolate measured temperature and precipitation across the country. These interpolated values were then used to calculate the other hydrological variables using the Waterloo Flood Forecasting Model (WATFLOOD). The Waterloo Mapping technique (WATMAP) was developed to use topographic and land cover databases to automatically and systematically derive the information needed to create the drainage database. WATFLOOD was calibrated with the Dynamically Dimensioned Search (DDS) algorithm using the difference between the measured and simulated streamflow as the objective function. After a final calibration of 100 separate DDS runs, distributed time series for the hydrological variables were created. A simple assessment was made for the predictive uncertainty in the simulated streamflow results based on the results of the final calibration. As well, the implications of various climate change scenarios were examined in the context of how they would change the hydrological variables. The major recommendations for future study included: finding other gridded datasets that could be used to verify the ones that were created in this study and examining further the magnitudes of the different kinds of predictive uncertainty (data, model, and parameter). The results of this thesis fit in well with the goals of the study on Predictions in Ungauged Basins. This thesis was organized along the principle of “design the process, not the product”. As such, although a set of final products are presented at the end, the most important part of the thesis was the process that achieved these products. Thus it is not assumed that every technique designed in this thesis will be applicable to every other researcher, but it hoped that most researchers in the field will be able to use at least some parts of the techniques developed here.

There is a lack of measured, long-term, reliable, and well-distributed hydrological variables in Canada. These hydrological variables include, but are not limited to: temperature, precipitation, ground runoff, evapotranspiration, soil moisture, and snow water equivalent. The objective of this thesis was to establish the best possible distributed estimates of these hydrological variables for Canada over the period of 1961-2000. The first step was to interpolate measured temperature and precipitation across the country. These interpolated values were then used to calculate the other hydrological variables using the Waterloo Flood Forecasting Model (WATFLOOD). The Waterloo Mapping technique (WATMAP) was developed to use topographic and land cover databases to automatically and systematically derive the information needed to create the drainage database. WATFLOOD was calibrated with the Dynamically Dimensioned Search (DDS) algorithm using the difference between the measured and simulated streamflow as the objective function. After a final calibration of 100 separate DDS runs, distributed time series for the hydrological variables were created. A simple assessment was made for the predictive uncertainty in the simulated streamflow results based on the results of the final calibration. As well, the implications of various climate change scenarios were examined in the context of how they would change the hydrological variables. The major recommendations for future study included: finding other gridded datasets that could be used to verify the ones that were created in this study and examining further the magnitudes of the different kinds of predictive uncertainty (data, model, and parameter). The results of this thesis fit in well with the goals of the study on Predictions in Ungauged Basins. This thesis was organized along the principle of “design the process, not the product”. As such, although a set of final products are presented at the end, the most important part of the thesis was the process that achieved these products. Thus it is not assumed that every technique designed in this thesis will be applicable to every other researcher, but it hoped that most researchers in the field will be able to use at least some parts of the techniques developed here.

This study aims to analysis the seasonal fluctuation fisheries assemblages and fishing yield by weight and amount of per unit effort using commercial shrimp beam trawlers at the waters off Yunlin, southwestern Taiwan during 1997-2007. In total, 47 sampling cruises, 320 taxa belonging to 205 genera and 102 families were identified. Among these, twelve species occurred greater than 80%. The species number of teleost fish were the most in each sampling. The total fish yield were 3032.9 kg and 304,680 individuals in the sampling period. Shrimps, the target species of the commercial beam trawler, were collected 945.7kg (31.2% of the total fish yield) and 181,050 individuals (59.4% of the total fish yield) , both are the most dominant fish yield among the fishing communities. The teleost fish (865.2 kg, 31.2%) and snails (73,868 individuals, 24.2%) were with the most fish yield in terms of weight and amount, respectively. The dominant species presented a seasonal pattern with a highly oscillation yield in both weight and biomass of the species in the same season each year. Thirteen species were recognized as the mainly harvested animals in of seasonal the commercial shrimp beam trawler. This study also showed that the shrimp, Parapenaeopsis hardiwckii, from February to May, and the portunid crab, Portunus sanguinolentus, from August to November in each year had the most production by cluster analysis. Since, the coastal zone at Yunlin were selected for constructing as an industrial park from May, 1998. Therefore in related to the development of the industrial park, to compare the production of 13 mainly harvest animals individually between three industrial developmental stages (pre-development: 1997.1-1998.2, development: 1998.5-2004.12 and post-development: 2005.2-2007.12). The results showed that the production of 3 species, Tanea lineate, Glossaulax didyma and Parapenaeopsis hardiwckii, decreased significantly (p< 0.05), where as of 2 fish species, Arius maculates and Chrysochir aureus, increased. By in large the other species was not remained the same after the construction.