The purpose of this work is to investigate methods of obtaining special percentage points of the studentized range, In fulfilling this purpose, two new methods are developed and used. / Master of Science

The problem discussed in this paper is the estimation of a single missing observation, two missing observations and several missing observations in a Group Visible (Singular) for Shirley balanced incomplete blocks design with two associate classes. Subsequently the analysis of variance, of the data augmented by the estimates of the missing observations, is derived. The method, first employed by Yates (1933), was followed to minimize the error sum of squares. Explicit formulae were developed, for the estimates of one missing observation, two missing observations occurring in various configurations and general formulae for z (= n) missing observations for certain particular configurations. Analysis of the data augmented by the estimates of the missing observations lead to positive bias in the case of treatment sum of squares, a method of analysis was discussed to eliminate this bias. A numerical example of illustrating the technique of estimating missing observations in a GDS P.B.I.B. design was given. The approximate and exact tests were performed, for the null hypothesis of no treatment differences, using the intra-block error mean square. / Master of Science

A method has been proposed for obtaining confidence intervals for the differences between treatment means in an analysis of variance. The intervals have the following properties: (1) The probability that the interval will cover the parameter is greater than or equal to (1 - α), and (2) If the same procedure is applied simultaneously to each of the pO2 differences among p means, the probability that all confidence intervals will cover the parameters correctly is at least (1 - α) ^{p - 1}. The same properties hold if the procedure is simultaneously applied to special linear comparisons among the means as well as to differences between single means. The intervals are complex in that the limits are dependent on the values of nuisance parameters. Three alternatives for handling these nuisance parameters are discussed, and one is preferred for use in practice. / Master of Science

Co-channel interference limits the capacity of cellular systems, reduces the throughput of wireless local area networks, and is the major hurdle in deployment of high altitude communication platforms. It is also a problem for systems operating in unlicensed bands such as the 2.4 GHz ISM band and for narrowband systems that have been overlaid with spread spectrum systems. In this work we have developed model-based techniques for the demodulation of co-channel MSK signals. It is shown that MSK signals can be written in the linear model form, hence a minimum variance unbiased (MVU) estimator exists that satisfies the Cramer-Rao lower bound (CRLB) with equality. This framework allows us to derive the best estimators for a single-user and a two-user case. These concepts can also be extended to wideband signals and it is shown that the MVU estimator for Direct Sequence Spread Spectrum signals is in fact a decorrelator-based multiuser detector. However, this simple linear representation does not always exist for continuous phase modulations. Furthermore, these linear estimators require perfect channel state information and phase synchronization at the receiver, which is not always implemented in wireless communication systems. To overcome these shortcomings of the linear estimation techniques, we employed an autoregressive modeling approach. It is well known that the AR model can accurately represent peaks in the spectrum and therefore can be used as a general FM demodulator. It does not require knowledge of the exact signal model or phase synchronization at the receiver. Since it is a non-coherent reception technique, its performance is compared to that of the limiter discriminator. Simulation results have shown that model-based demodulators can give significant gains for certain phase and frequency offsets between the desired signal and an interferer. / Master of Science

No / This paper employs the standard General Auto-regressive Conditional Heteroskedasticity (GARCH(1,1)) process to examine the impact of option listing on volatility the underlying stocks. It takes into consideration the time variation in the individual stock's variance and explicitly tests whether option listing causes any permanent volatility change. It also investigates the impact of option listing on the speed at which information is incorporated into the stock price. The study uses clean samples to avoid sample selection biases and control samples to account for the change in the volatility and/or information flows that may be caused by factors other than option listing.

This thesis will investigate if adding an exogenous variable (implied volatility) to the variance equation will increase the performance for the GARCH(1,1) and EGARCH(1,1) models based on the OMXS30 index. These models are also compared with the implied volatility itself as a forecasting/modeling method. To evaluate the models the realized variance will be used as an unbiased estimator of the conditional variance. The findings suggest that adding implied volatility to the variance equation increase the overall performance.

La présente thèse s'est inscrite dans une perspective d'améliorer les outils de mesure du risque de marché en proposant des solutions capables de reproduire certaines caractéristiques empiriques d'évolution des marchés financiers. A travers une étude empirique sur des données réelles, nous montrons que la réalité des marchés financiers possède certaines caractéristiques empiriques connues et résumées sous le nom "faits stylisés", qui rendent les mesures usuelles du risque de marché incapables de reproduire ces caractéristiques. Nous proposons des nouvelles méthodes de mesure de la Value-at-Risk (VaR), en fonction de la volatilité passée et des corrélations existant entre les actifs composant un portefeuille, dans le cadre de deux grandes approches de mesure du risque : une approche de mesure du risque global (ou risque univarié) et une approche de mesure du risque multiple (ou risque multivarié), tout en testant leur qualité prédictive au moyen des procédures de backtesting. Les résultats obtenus montrent une grande capacité des différentes mesures utilisées à capturer les faits stylisés caractérisant l'évolution des marchés financiers étudiés avec une nette surperformance des méthodes de mesure de la VaR estimées dans le cadre du risque multivarié par rapport à celles du risque univarié. / This thesis has object to improve the methods for estimating market risk by offering solutions capable to replicate some empirical properties of asset returns. Through an empirical study on real data, we show that the reality of financial markets has some empirical characteristics known and summarized as "stylized facts" that render the conventional market risk measurement unable to reproduce. We propose a Value-at-Risk (VaR) measures, based on modeling portfolio volatility and correlations between assets classes, using two risk measurement approaches: an univariate risk measurement approach and multivariate risk measurement approach, and testing their quality predictive using backtesting procedures. The results obtained show a great ability of different used risk measurement to capture the stylized facts characterizing financial markets, with a clear outperformance of the multivariate VaR measures than the univariate VaR measures.

This thesis focuses on impact of jumps and simultaneous jumps (co-jumps) in asset prices on future volatility. Our main contribution to the empirical literature lies in the use of panel Heterogeneous Autoregressive (HAR) model that allows us to obtain average effect of jumps for both the portfolio of 29 U.S. stocks and 8 individual market sectors our stocks belong to. On top of that we investigate the effect of sign for both jumps and co-jumps. The estimation results indicate that the impact of jumps on future volatility is positive whereas for co-jumps it is negative. We also document tendency of downward jumps and co-jumps to be followed by increase in volatility and that upward jumps and co-jumps are followed by decrease in volatility. Finally, results for individual sectors reveal that estimated effects vary across industries - for cyclical sectors volatility is in general more sensitive to negative jumps and less sensitive to positive jumps than for defensive sectors.

Les systèmes mécaniques soumis et excités par vibrations sont les candidats naturels à être modélisé par des systèmes linéaires invariables dans le temps. La localisation de dommages utilisant les paramètres modaux évalués à partir de données de vibration ambiantes mesurées grâce à de capteurs est possible notamment par l'approche nommée Stochastic Dynamic Damage Location Vector (SDDLV), où l'emplacement des dommages est empiriquement relié aux positions où le stress est proche de zéro. La première contribution dans cette thèse montre comment l'incertitude sur les paramètres du système d'état peut être utilisée pour déduire des bornes d'incertitude sur les résidus de localisation de dommages, ceci afin de décider de l'emplacement de dommage utilisant un test d'hypothèse. Dans la deuxième contribution, la méthode de localisation de dommages est étendue pour être robuste au choix des variables de Laplace utilisées dans cette méthode. Ceci est obtenue en agrégeant statistiquement les résultats à valeurs différentes dans le domaine de Laplace. L'influence Line Damage Location (ILDL) est une approche complémentaire du SDDLV où l'angle entre les sous-espaces principaux est calculé et les dommages sont empiriquement localisés aux points près du zéro. L'approche développée pour la SDDLV est étendue à cette nouvelle approche, l'ILDL. Les méthodes proposées sont validées et appliquées avec succès pour la localisation de dommages dans des structures civiles. / Mechanical systems under vibration excitation are prime candidate for being modeled by linear time invariant systems. Damage localization using both finite element information and modal parameters estimated from ambient vibration data collected from sensors is possible by the Stochastic Dynamic Damage Location Vector (SDDLV) approach, where the damage location is empirically related to positions where the stress is close to zero. The first contribution in this thesis shows how the uncertainty in the estimates of the state space system can be used to derive uncertainty bounds on the damage localization residuals to decide about the damage location with a hypothesis test using one chosen Laplace value. In the second contribution, the damage localization method is extended with a statistical framework and robustness of the localization information is achieved by aggregating results at different values in the Laplace domain. The Influence Line Damage Location (ILDL) is a complementary approach of the SDDLV where the subspace angle is computed and damage is empirically located at points near zero. The last contribution describes how robustness of the localization information is achieved by aggregating results at different values in the Laplace domain based on the previous two contributions. The proposed methods are validated and successfully applied to damage localization of several applications in civil structures.

This study sought to determine the effect upon Tukey's Honestly Significant Difference (HSD) statistic of concurrently violating the assumptions of homogeneity of variance and equal sample sizes. Two forms for the unequal sample size problem were investigated. Kramer's form and the harmonic mean approach were the two unequal sample size procedures studied. The study employed a Monte Carlo simulation procedure which varied sample sizes with a heterogeneity of variance condition. Four thousand experiments were generated. Findings of this study were based upon the empirically obtained significance levels. Five conclusions were reached in this study. The first conclusion was that for the conditions of this study the Kramer form of the HSD statistic is not robust at the .05 or .01 nominal level of significance. A second conclusion was that the harmonic mean form of the HSD statistic is not robust at the .05 and .01 nominal level of significance. A general conclusion reached from all the findings formed the third conclusion. It was that the Kramer form of the HSD test is the preferred procedure under combined assumption violations of variance heterogeneity and unequal sample sizes. Two additional conclusions are based on related findings. The fourth conclusion was that for the combined assumption violations in this study, the actual significance levels (probability levels) were less-than the nominal significance levels when the magnitude of the unequal variances were positively related to the magnitude of the unequal sample sizes. The fifth and last conclusion was that for the concurrent assumption violation of variance heterogeneity and unequal sample sizes, the actual significance levels significantly exceed the nominal significance levels when the magnitude of the unequal variances are negatively related to the magnitude of the unequal sample sizes.