Submitted by Ludmilla Oliveira Ambrosi Sales (ludy.sales@gmail.com) on 2017-02-19T02:58:28Z No. of bitstreams: 1 Dissertação de Ludmila_FGV.pdf: 1634569 bytes, checksum: f69fd3c3e31851a5d2a8496dbd9c50a8 (MD5) / Rejected by Renata de Souza Nascimento (renata.souza@fgv.br), reason: Trabalho submetido duas vezes. on 2017-02-20T16:33:52Z (GMT) / Submitted by Ludmilla Oliveira Ambrosi Sales (ludy.sales@gmail.com) on 2017-02-20T21:43:17Z No. of bitstreams: 1 Dissertação de Ludmila_FGV_.pdf: 2700997 bytes, checksum: 502da2dea23764b52c65a0ab70a00a9f (MD5) / Rejected by Renata de Souza Nascimento (renata.souza@fgv.br), reason: Ludmilla, Está correto, porém, o código da ficha catalográfica (CDU 336.76) deve estar ao lado direito da ficha. Aguardo on 2017-02-20T21:48:45Z (GMT) / Submitted by Ludmilla Oliveira Ambrosi Sales (ludy.sales@gmail.com) on 2017-02-20T23:06:20Z No. of bitstreams: 1 Dissertação de Ludmila_FGV_.pdf: 2701182 bytes, checksum: c86ac2ab833162046024483778a8b39a (MD5) / Approved for entry into archive by Renata de Souza Nascimento (renata.souza@fgv.br) on 2017-02-20T23:28:14Z (GMT) No. of bitstreams: 1 Dissertação de Ludmila_FGV_.pdf: 2701182 bytes, checksum: c86ac2ab833162046024483778a8b39a (MD5) / Made available in DSpace on 2017-02-21T18:12:36Z (GMT). No. of bitstreams: 1 Dissertação de Ludmila_FGV_.pdf: 2701182 bytes, checksum: c86ac2ab833162046024483778a8b39a (MD5) Previous issue date: 2017-01-27 / This paper revisits the theory of market efficiency and analyzes the Brazilian capital market for a more recent period in order to verify if the improvement pointed out in the study by Bonomo (2002) persists, that is, if the reduction of inefficiency in the course of the Time is robust. The existence of autocorrelation may be an indication of abnormal returns if the strategies adopted exploit this correlation and generate an abnormal return. The autocorrelation tests adopted in the random walk literature, for the most part, do not take into account the Heteroscedasticity characteristic of financial assets and, therefore, this work seeks to apply Bartlett’s formula for non-linear processes in order to verify if existence Of autocorrelation between the Brazilian papers analyzed and if this is enough to generate an extraordinary return. Traditional statistical and correlation tests were applied together with random walk tests to verify if the Brazilian capital market is efficient in its weak form. / Este trabalho revisita a teoria de eficiência de mercado e analisa o mercado de capitais brasileiros para um período mais recente a fim de verificar se a melhora apontada no estudo feito por Bonomo (2002) persiste, ou seja, se a redução da ineficiência no decorrer do tempo é robusta. Foram selecionadas 15 ações brasileiras que compunham o IBOVESPA de Maio 2016 e o período de análise compreende Janeiro de 2000 a Maio 2016. A existência de autocorrelação pode ser um indício de retornos anormais caso as estratégias adotadas explorem essa correlação e consigam gerar um retorno anormal. Os testes de autocorrelação adotados na literatura de passeio aleatório, em sua maioria, não levam em conta a característica de Heterocedasticidade dos ativos financeiros e, por isso, este trabalho busca aplicar a fórmula de Bartlett para processos não lineares a fim de verificar se a existência de autocorrelação entre os papéis brasileiros analisados e se esta é suficiente para gerar um retorno extraordinário. Testes estatísticos tradicionais e de correlação foram aplicados juntamente a testes de random walk para verificar se o mercado de capitais brasileiro é eficiente na sua forma fraca.

Valutamarknaden är världens största marknad och en nödvändig del av dagens globala samhälle, som gör det möjligt för företag att göra affärer i olika valutor och mellan olika gränser. Marknaden utgör en stor handelsplattform för både små och stora aktörer, för vilka det är viktigt att prognostisera valutakurser med gott resultat. Att modellera finansiella instrument i form av tidsserier är en av de vanligaste investeringsstrategierna och dess användningsområde sträcker sig från valutamarknaden till bland annat aktiemarknaden och råvarumarknaden. I denna uppsats undersöks fyra olika statistiska metoder för att modellera valutakursen Euro-US Dollar givet historisk data, och prognoser görs med de framtagna modellerna. Dessa metoder är slumpvandring, ARIMA, ARIMA-GARCH och VAR. Vidare undersöks för den dynamiska VAR-modellen hur valutamarkanden påverkar, och blir påverkad av, långa och korta räntan. Resultaten visar att ARIMA(3,1,2) förklarar valutakursen bäst medan VAR(2) med valutakursen och skillnaden mellan långa räntor som ingående variabler ger de bästa prediktionerna. / The foreign exchange market is the world’s largest market and is an essential part of the global society of today. The FX market enables companies to trade with different currencies across country borders. It is also a large trade-platform for both big and small financial actors, who greatly benefit from the advantages of good predictions. Modeling of financial instruments is one of the most commonly used investment strategies and its area of application ranges from the FX market to markets suchas the stock market and the commodity market. In this paper, four different statistical models are used to model the currency pair Euro-US Dollar. These methods are random walk, ARIMA, ARIMA-GARCH and VAR. Besides investigating which method that gives the best forecasts, the method that best describes the training datais also found. Furthermore, for the dynamic VAR model, it is explored how the FX market affects, and is affected by, the long term and short term interest. The results show that ARIMA(3,1,2) is the best at describing the exchange rate while VAR(2) with the exchange rate and the difference between long term interests as variables gives the best predictions.

Förvärv av företag sker till höger och vänster, i alla länder, branscher och kategorier. Motiven till förvärv må vara av varierande art, men det som är av betydelse för ett företag är vilket värde ett förvärv kan generera. Denna studie undersöker marknadsreaktioner vid olika typer och kombinationer av förvärv; däribland horisontella, vertikala och branschspecifika förvärv, med huvudsaklig inriktning på värdet som skapas vid gränsöverskridande och nationella förvärv. Detta görs i syfte att kartlägga eventuella överlägsna typer av förvärv. Klassiska teorier i form av effektiva marknadshypotesen och “random walk”-teorin i kombination med insiderhandel utgör den vetenskapliga utgångspunkten i studien. Metoden är en eventstudie, där den abnormala avkastningen vid dessa förvärv beräknas, för att kunna analyseras i förhållande till tidigare forskning. Vidare testas studiens hypoteser med hypotesprövning och samband analyseras med hjälp av enkla linjära regressioner som sedan tolkas. Resultatet visar på att gränsöverskridande förvärv ger den högsta genomsnittliga kumulativa abnormala avkastningen (CAAR), och vidare är horisontella förvärv att föredra över vertikala av samma anledning. Offentliga sektorn och tjänstesektorn ger de högsta CAAR sett till branscher. / Merger and acquisitions is happening everywhere; in all countries, industries and in all categories. The motives of acquiring other companies may differ, but what is truly meaningful is the value an acquisition can bring. This study deals with different types and combinations of acquisitions; amongst horizontal, vertical and industry specific acquisitions, with main focus on the value created in Cross Border and national acquisitions. The aim of the study is thereby to plot superior types of acquisitions. Classic theories like the Efficient market hypothesis and the Random walk theory will be used as a theoretical framework, along with the idea of Insider trading. The method used is an Event study, where the abnormal return is calculated at the announcement day of acquisition and is later analyzed in association to prior studies. Deduced hypothesis are then processed in hypothesis testing and correlations are being studied through simple linear regressions. The result show that Cross Border acquisitions give the highest value (CAAR) to acquirer and the horizontal acquisitions are superior to vertical ones for the same reason. The public sector and the service sector yield the highest CAAR when industries are being studied.

Les gènes et les protéines n’agissent pas de manière isolée dans les cellules, mais interagissent plutôt pour faire leurs fonctions dans les processus biologiques. Ces interactions peuvent être représentées sous forme de grands réseaux dans lesquels les nœuds sont des gènes ou des protéines et les arêtes représentent leurs interactions. Diverses approches basées sur la théorie des graphes ont été développées pour extraire la connaissance fonctionnelle contenue dans ces réseaux. Néanmoins, ces méthodes ont été principalement appliquées à des réseaux individuels, en ignorant la diversité des interactions biologiques. Nous déclarons que ces différents types d’interactions peuvent être représentés sous la forme de réseaux multiplexes, c’est-à-dire des ensembles de réseaux partageant les mêmes nœuds, ce qui permet une description plus précise des systèmes biologiques. Cette thèse est focalisée sur le développement de nouveaux algorithmes étendant aux réseaux multiplexes certaines méthodes populaires de la théorie des graphes en biologie computationnelle, ainsi que sur leur application à l’étude des maladies humaines. Du côté des applications, nous nous concentrons sur les maladies liées au vieillissement prématuré, un groupe de maladies génétiques ressemblant à certains aspects du vieillissement physiologique à un âge précoce. Nous avons appliqué nos algorithmes pour détecter les modules associés à plus de 70 syndromes annotés avec un phénotype lié au vieillissement prématuré. Les résultats ont révélé le paysage des processus moléculaires perturbés dans ces maladies, qui peuvent être mis en parallèle avec les caractéristiques du vieillissement physiologique. / Genes and proteins do not act isolated in cells but rather interact to perform their functions in signaling pathways, molecular complexes, or, more generally, biological processes. These interactions can be represented as large networks in which nodes are genes or proteins and edges represent their interactions. Various graph-theory based approaches have been developed to extract the functional knowledge contained in biological networks. Nevertheless, these methods have been mainly applied to individual networks, ignoring the diversity of biological interactions. We state here that these different types of interactions can be represented as multiplex networks, i.e. collections of networks sharing the same nodes, leading to a more accurate description of biological systems. This thesis focuses on the extension from individual to multiplex networks of some of the state-of-the-art guilt-by-association methods in computational biology, and on their application to the study of human diseases. On the application side, we concentrate on premature aging diseases, a group of rare genetic disorders that resemble some aspects of physiological aging at an early age. In this framework, we applied our algorithms to detect the modules associated to more than 70 disorders annotated with at least one premature aging related phenotype. The results revealed the landscape of perturbed molecular processes in premature aging diseases, which can be paralleled with the hallmarks of physiological aging to help identifying common and specific features.

This essay examines the impact of game releases on the Swedish stock market. As previous research has examined product launches and news releases, this thesis intends to investigate game releases by game developer companies such as developers of computer, console or mobile games. Previous research has been based on a business perspective and business valuation, the authors of this thesis intend to examine the individual investor's perspective. The theoretical framework consists of information asymmetry, the signaling model, the effective market hypothesis, random walk hypothesis and market reaction to new products. The study has chosen a deductive quantitative research approach with the event study method. The sample consists of 14 game developer companies in computer, console and mobile games during theperiod 2017–2021 that are listed on the Swedish market, which were observed during a period of 180 days before the event day and 40 days after. The results showed that there is a statistically significant relationship before, after and during the event day for game releases.The period during and after the event day can also be generalized where 9/14 respective 10/14 game developer companies showed statistically significant within the accumulated abnormal return. / Denna uppsats undersöker påverkan av spelsläpp på den svenska aktiemarknaden. Då tidigare forskning har undersökt produktlanseringar och nyhetssläpp ämnar denna uppsats att undersöka spelsläpp av spelutvecklarebolag som utvecklar PC-, konsol eller mobilspel. Tidigare forskning har utgått från ett företagsperspektiv och företagsvärdering, författarna för denna uppsats ämnar undersöka den individuella investerarens perspektiv. Det teoretiska ramverket består av informationsasymmetri, signaleringsmodellen, den effektiva marknadshypotesen, random walk hypothesis och marknadsreaktion till nya produkter. Studien har en deduktiv kvantitativa forskningsansats med eventstudie metoden. Urvalet består av 14 spelutvecklarebolag inom dator-, konsol- och mobilspel under tidsperioden 2017–2021 som är börsnoterade på den svenska marknaden. Dessa bolag observerades under en tidsperiod 180 dagar innan eventdagen och 40 dagar efter. Resultatet visade på att det finns ett statistiskt signifikant samband innan, efter och under eventdagen för spelsläpp. Perioden under och efter eventdagen kan även generaliseras där 9/14 respektive 10/14 spelutvecklarbolag visade på statistisk signifikant inom den ackumulerade abnormala avkastning.

<p>In recent years there has been a growing interest in data mining and graph machine learning for techniques that can obtain frequencies of <em>k</em>-node Connected Induced Subgraphs (<em>k</em>-CIS) contained in large real-world graphs. While recent work has shown that 5-CISs can be counted exactly, no exact polynomial-time algorithms are known that solve this task for <em>k </em>> 5. In the past, sampling-based algorithms that work well in moderately-sized graphs for <em>k</em> ≤ 8 have been proposed. In this thesis I push this boundary up to <em>k</em> ≤ 16 for graphs containing up to 120M edges, and to <em>k</em> ≤ 25 for smaller graphs containing between a million to 20M edges. I do so by re-imagining two older, but elegant and memory-efficient algorithms -- FANMOD and PSRW -- which have large estimation errors by modern standards. This is because FANMOD produces highly correlated k-CIS samples and the cost of sampling the PSRW Markov chain becomes prohibitively expensive for k-CIS’s larger than <em>k </em>> 8.</p> <p>In this thesis, I introduce:</p> <p>(a)  <strong>RTS:</strong> a novel regenerative Markov chain Monte Carlo (MCMC) sampling procedure on the tree, generated on-the-fly by the FANMOD algorithm. RTS is able to run on multiple cores and multiple machines (embarrassingly parallel) and compute confidence intervals of estimates, all this while preserving the memory-efficient nature of FANMOD. RTS is thus able to estimate subgraph statistics for <em>k</em> ≤ 16 for larger graphs containing up to 120M edges, and for <em>k</em> ≤ 25 for smaller graphs containing between a million to 20M edges.</p> <p>(b) <strong>R-PSRW:</strong> which scales the PSRW algorithm to larger CIS-sizes using a rejection sampling procedure to efficiently sample transitions from the PSRW Markov chain. R-PSRW matches RTS in terms of scaling to larger CIS sizes.</p> <p>(c) <strong>Ripple:</strong> which achieves unprecedented scalability by stratifying the R-PSRW Markov chain state-space into ordered strata via a new technique that I call <em>sequential stratified regeneration</em>. I show that the Ripple estimator is consistent, highly parallelizable, and scales well. Ripple is able to <em>count</em> CISs of size up to <em>k </em>≤ 12 in real world graphs containing up to 120M edges.</p> <p>My empirical results show that the proposed methods offer a considerable improvement over the state-of-the-art. Moreover my methods are able to run at a scale that has been considered unreachable until now, not only by prior MCMC-based methods but also by other sampling approaches. </p> <p><strong>Optimization of Restricted Boltzmann Machines. </strong>In addition, I also propose a regenerative transformation of MCMC samplers of Restricted Boltzmann Machines RBMs. My approach, Markov Chain Las Vegas (MCLV) gives statistical guarantees in exchange for random running times. MCLV uses a stopping set built from the training data and has a maximum number of Markov chain step-count <em>K</em> (referred as MCLV-<em>K</em>). I present a MCLV-<em>K</em> gradient estimator (LVS-<em>K</em>) for RBMs and explore the correspondence and differences between LVS-<em>K</em> and Contrastive Divergence (CD-<em>K</em>). LVS-<em>K</em> significantly outperforms CD-<em>K</em> in the task of training RBMs over the MNIST dataset, indicating MCLV to be a promising direction in learning generative models.</p>

The study investigates the presence of the momentum effect in the Swedish stock market by utilizing both cross-sectional introduced by Jegadeesh and Titman (1993) and time-series momentum introduced by Moskowtozt et al. (2011). The period of analysis is between 1998 to 2022. Additionally, the study compares the performance of these two momentum strategies by creating portfolios with varying lookback and holding periods. However, the primary focus is on the strategy with a 12-month lookback and a 1-month holding period. The results indicate that both momentum strategies generated positive returns over the analyzed period. However, time-series momentum was more effective for longer lookback periods, while cross-sectional momentum was more effective for shorter periods. Nevertheless, none of the findings for either momentum strategy were statistically significant in the Swedish stock market. / Syftet med denna studie är att studera lönsamheten för momentumstrategier på den svenska marknaden för perioden januari 1998 till december 2022, med hjälp av de cross-sectional och time-series momentum som introducerades av Jegadeesh och Titman (1993) och Moskowitz et al. 2011), respektive. Resultaten visar att momentumstrategier har positiv avkastning på den svenska marknaden, men resultaten är inte statistiskt signifikanta. Under den tidsperiod under vilken momentumstrategier testades visade cross-sectional momentum bättre resultat under kortare utvärderingsperioder jämfört med time-series momentum, som presterade bättre under längre utvärderingsperioder. Det är dock värt att notera att även om momentumportföljers positiva avkastning inte är statistiskt signifikanta, kan de fortfarande vara fördelaktiga för avkastningssökande investerare. Dessutom fann studien att den positiva avkastningen inte enbart beror på momentumfaktorn utan kan också bero på portföljens exponering mot Fama French SMB-faktorn.

共三十冊，三萬五千字，分四章。 在古典學派的消費函數理論，消費係利率的函數，到了凱因斯，所得才正式出現在消 費函數，由於凱因斯對消費函數的設定，係依所謂基本心理法則而來，乃引起學者研 究消費函數的興趣，各種學說紛紛出現，乃形成百家爭鳴的局面。在各種理論中，以 M．Friendman 和 A ．Modigliani顯得較為突出，在此兩家學說中，雖然引進預期所 得的概念，然而他們係以過去的所得來估計未來的預期所得。本文試著引進理性預期 的概念入未來預期的所得，亦即假設個人對未來所得的預期是理性的。在這樣前提下 ，再加上M．Friedman 的消費理論，吾人可得出消費呈Random walk 的形式。最後吾 人利用台灣資料來測試消費是否呈 Randon walk。

建立模型來估計匯率早已行之有年。較早期的匯率模型，不論是在樣本內的配適或是樣本外的預測，其實表現的並不理想。之後的研究針對這樣的結果指出，這是因為匯率的表現是非線性的，並非傳統線性模型可描繪出來。而對於捕捉匯率非線性的特性，傾向使用無母數的估計方式。因此，本研究採用最近鄰域法進行美元兌新台幣的匯率預測。另外，許多早期的研究發現，隨機漫步模型與其他模型相比較之後，在匯率預測上的表現最好，因而引發了”打敗隨機漫步”的一連串熱潮。本研究欲延續這項議題，將隨機漫步模型做為與最近鄰域模型比較的基準。 / 本研究使用的資料為即期匯率，包含日資料、週資料和月資料三種。將每種資料皆切割為樣本內與樣本外兩個部分，其中最後三分之一的樣本數用於樣本外預測。平均絕對誤差與平均誤差平方根則是用來衡量比較模型預測的準確性。實證結果發現，使用局部加權估計的最近鄰域模型在樣本內的配適表現上優於隨機漫步模型；然而，在樣本外的預測能力上，隨機漫步模型仍舊略勝一籌。 / A wide variety of empirical exchange rate models have been estimated over the years. Earlier findings indicated that exchange rate equations do not fit particularly well, and forecast no better. Later researches then provided a potential reason for the poor performance that traditional exchange rate models, because they are nonlinear. To find a resolution for nonlinearity, nonparametric techniques tend to be useful tools. In this study, we use one of nonparametric techniques called nearest-neighbors method to predict NTD against USD. Besides, many earlier papers found that forecasts from popular models for the foreign exchange rate generally fail to improve upon the random walk out-of-sample. “Beat the random walk” became an emerging issue then. This has motivated this research, and thus we include the random walk as a linear benchmark. / The data set consists of the daily, weekly and monthly spot rates for NTD/USD. We divide each data set into a fitting set and a prediction set for in-sample analysis and out-of-sample forecast, respectively. The out-of-sample forecasts are calculated from the last one-third of each series. As a measure of performance the mean squared error (MAE) and root mean squared error (RMSE) are used. In our empirical results, we find that nearest-neighbors model using local weights easily tops the random walk in-sample. However, as we turn to the out-of-sample prediction, no models produce forecasts superior to the random walk. It seems difficult to beat the random walk out-of-sample in this study.

In an effort to assess the predictive ability of exchange rate models when data on African countries is sampled, this paper studies nonlinear modelling and prediction of the nominal exchange rate series of the United States dollar to currencies of thirty-eight African states using the smooth transition autoregressive (STAR) model. A three step analysis is undertaken. One, it investigates nonlinearity in all nominal exchange rate series examined using a chain of credible statistical in-sample tests. Significantly, evidence of nonlinear exponential STAR (ESTAR) dynamics is detected across all series. Two, linear models are provided another chance to make it right by shuffling to data on African countries to investigate their predictive power against the tough random walk without drift model. Linear models again failed significantly. Lastly, the predictive ability of nonlinear models against both the random walk without drift and the corresponding linear models is investigated. Nonlinear models display useful forecasting gains over all contending models.