Direct marketing problems have been intensively reviewed in the marketing literature recently, such as purchase frequency and time, sales profit, and brand choices. However, modeling the customer response, which is an important issue in direct marketing research, remains a significant challenge. This thesis is an empirical study of predicting customer response to direct marketing and applies a Bayesian approach, including the Bayesian Binary Regression (BBR) and the Hierarchical Bayes (HB). Other classical methods, such as Logistic Regression and Latent Class Analysis (LCA), have been conducted for the purpose of comparison. The results of comparing the performance of all these techniques suggest that the Bayesian methods are more appropriate in predicting direct marketing customer responses. Specifically, when customers are analyzed as a whole group, the Bayesian Binary Regression (BBR) has greater predictive accuracy than Logistic Regression. When we consider customer heterogeneity, the Hierarchical Bayes (HB) models, which use demographic and geographic variables for clustering, do not match the performance of Latent Class Analysis (LCA). Further analyses indicate that when latent variables are used for clustering, the Hierarchical Bayes (HB) approach has the highest predictive accuracy.

Advances in computing has undoubtfully been one of the main catalysts in the formation of the discipline always known as Statistics. A fundamental question addressed here is whether computing facilities, such as parallel or high performance computing, could assist in the development of methodologies that render stronger results, based on some predetermined optimality criterion. The candidate at the hand of which this enquiry is made, is the arc length of some statistical function. Estimation, goodness-of-fit, linear regression and non-linear regression, which may all be considered as central themes in Statistics, are revisited, and redefined in terms of this new measure. The results resulting from these arc length methodologies are obtained from simulation, as well as from real case studies, and contrasted to that obtained using their classical counterparts. Mathematical premises for the proposed methods are provided, together with the documentation accompanying the companion R package, along with the data utilised for the applications. / Thesis (PhD)--University of Pretoria, 2017. / National Research Foundation of South Africa, Unique Grant No. 94108. / Statistics / PhD / Unrestricted

Diploma thesis "Mediation Analysis in Sociology" deals with mediation analysis and possibilities of its application in sociology, depending on the type of the dependent variable that enters the analysis. In the first case the dependent variable is continuous - in this case the SPSS software and its PROCESS add-on are used to directly analyse the mediation. In the second case the dependent variable that enters the analysis is binary - the PROCESS add-on doesn't allow this option; therefore, the analysis is performed in SPSS software by the set of linear and logistic regressions according to the Baron & Kenny method. Two case studies from the field of sociology, GSS (General Social Survey) and ISSP (International Social Survey Programme), are used in the thesis and the consequences of the transition from continuous dependent variable to binary are examined using the secondary analysis of these data.

In practice, when one has many candidate variables as explanatory variables in multiple regression, there is always the possibility that variables that are important determinants of the response variable might be omitted from the model, while unimportant variables might be included. Both types of errors are important, and in this dissertation it is attempted to quantify the probabilities of these errors. A simulation study is reported in this dissertation. Different numbers of variables, i.e. p= 4 to 20 are assumed, and different sample sizes, i.e. n=0.5p, p, 2p, 4p. For each p the underlying model assumes that roughly half of the independent variables are actually correlated with the dependant variable and the other half not. The noise is ε~ N(0, σ2, where σ2, is set fixed. The data was simulated 10000 times for each combination of n and p using known underlying models and ε randomly selected from of a normal distribution. For this investigation the full model and forward selection regression are compared. The mean squared error of the estimated coefficient β(p) is determined from the true β of each n and p set. A full discussion, as well as graphs, is presented. / Dissertation (MSc)--University of Pretoria, 2011. / Statistics / unrestricted

As an investor, the goal is to get as high return at as low risk as possible on an invested stock. Theories show a positive relationship between high risk and high return. However, there are studies that contradict this statement, which creates misleading investors. The purpose of this quantitative study was to investigate the relationship between risk and return, to see if increased risk really leads to a higher return. The study is limited to the Frankfurt Stock Exchange between the period 2010-2019. The index examined is Germany's largest index DAX, which comprised 30 shares on 1 January 2015. Four portfolios have been created: two low-risk portfolios and two high-risk portfolios, where fictitious investments were made to then measure the return after a period of five years. The two risk measures used in the study for calculation are beta value and volatility. The beta value has been used in CAPM to calculate the expected return. The results of the study using correlation and multiple regression analysis obtained a negative significant relationship between risk and return. The portfolios also showed a negative relationship between higher risk and higher return. The results from this study are consistent with many previous studies, but the variables used to measure and define risk can be questioned. / För en investerare är målet att få en så hög avkastning till en så låg risk som möjligt. Finansiella teorier och allmän uppfattning hävdar att det finns ett positivt samband mellan hög risk och hög avkastning. Dock finns det mycket tidigare forskning som motstrider detta påstående, vilket kan bli vilseledande för investerare. Syftet med denna kvantitativa studie var att undersöka sambandet mellan risk och avkastning, för att se om ökad risk leder till en högre avkastning. Utöver det undersöktes om CAPM kunde ge en korrekt estimering av framtida förväntad avkastning. Studien avgränsades till Frankfurtbörsen mellan perioden 2010-2019. Det indexet som undersöktes var Tysklands största index DAX som det såg ut 2015-01-01, det omfattades då av 30 aktier. Fyra portföljer har skapats: två portföljer med låg risk och två portföljer med hög risk, där fiktiva investeringar gjordes för att sedan mäta avkastningen efter en tidsperiod på fem år. De två mått på risk som studien använt är betavärde och volatilitet. Betavärdet har använts i CAPM för att beräkna förväntad avkastning. Studiens resultat har med hjälp av korrelation och multipel regressionsanalys visat på ett negativt signifikant samband mellan risk och avkastning. Portföljerna visade också på ett negativt samband mellan högre risk och högre avkastning. Resultatet från denna studie är i enhetlighet med många tidigare studier, dock kan de variabler som använts för att mäta och definiera risk ifrågasättas.

The application area of Fifth Generation New Radio (5G-NR) called Ultra-Reliable and Low-Latency Communication (URLLC) requires a reliability, the probability of receiving and decoding a data packet correctly, of 1 - 10^5. For this requirement to be fulfilled in a resource-efficient manner, it is necessary to have a good estimation of extremely low quan- tiles of the channel quality distribution, so that appropriate resources can be distributed to users of the network system.  This study proposes and evaluates two methods for estimating extreme quantiles of the downlink channel quality distribution, linear quantile regression and Quantile Regression Neural Network (QRNN). The models were trained on data from Ericsson’s system-level radio network simulator, and evaluated on goodness of fit and resourcefulness. The focus of this study was to estimate the quantiles 10^2, 10^3 and 10^4 of the distribution.  The results show that QRNN generally performs better than linear quantile regression in terms of pseudoR2, which indicates goodness of fit, when the sample size is larger. How- ever, linear quantile regression was more effective for smaller sample sizes. Both models showed difficulty estimating the most extreme quantiles. The less extreme quantile to esti- mate, the better was the resulting pseudoR2-score. For the largest sample size, the resulting pseudoR2-scores of the QRNN was 0.20, 0.12 and 0.07, and the scores of linear quantile regression was 0.16, 0.10 and 0.07 for the respective quantiles 10^2, 10^3 and 10^4.  It was shown that both evaluated models were significantly more resourceful than us- ing the average of the 50 last measures of channel quality subtracted with a fixed back-off value as a predictor. QRNN had the most optimistic predictions. If using the QRNN, theo- retically, on average 43% more data could be transmitted while fulfilling the same reliability requirement than by using the fixed back-off value.

This thesis aims to investigate if the hourly heat load consumptiondata can be used to approximate the daily consumptions patterns forbuildings connected to Göteborg Energi’s district heating network. Theapproximated data shall act as foundation for dimensioning of thedistrict heating network. In this work, it is studied how theconsumption approximation are due to changes in the outdoortemperature between different years.The aim is to develop an approximation model for hourly heat loadpatterns, heat output, water flow and return temperature from thedistrict heating substations of individual buildings regardless ofbuilding types. The approximation methods used in the hourlyapproximation model is multiple ridge regression. Regression trees areused to define breaking points such as the building balance pointtemperature from the consumer heat load pattern. Two separateregression intervals were defined based on breaking points from theregression tree. Outdoor temperature data, solar radiation data,weekday and weekends data used as predictors.The approximation model is evaluated against a reference model usingthe daily mean heat load consumption data. Evaluation between themodel and reference is made on six different building and buildingtypes during the outdoor temperature of -16, which is the designoutdoor temperature of the district heating system of Göteborg Energi.The approximated maximum heat output and water flow during the daywhere 18 % and 10 % above the approximated daily mean. Theapproximated return temperature where 43-51 °C compared to the dailymean of 42 °C for a warm year and 47-52,5 °C compared to 50 °C dailymean for a cold year.The hourly approximation model where able to capture the heat loadpatterns of different building types. However, higher demands on dataquality needs to be addressed to ensure the use of the hourlyapproximation model. / I detta examensarbete har en undersökning angående värmelastapproximationer baserade påtimupplöst kundlastdata gjorts. Värmelasterna som approximerades var värmeeffekt,vattenflöde och returtemperatur. Data för utomhustemperatur, helg och vardag samtsolinstrålningsdata har använts för att kunna approximera värmelasterna. Resultat avapproximationer har visualiserats i relation till utomhustemperaturen och har utvärderats fördimensionerande utomhustemperatur. Utvärdering gjordes på olika byggnader ochbyggnadstyper. Resultat av approximationsmodell med timupplöst kundlastdata utvärderadesmot modell baserad på dygnsmedeldata. Modellerna testades för två olika år med skildautomhustemperaturer, ett kall-år och ett varm-år.Resultat visar att det är möjligt att fånga den timvisa värmelasten hos enskilda kunder och skulleinnebära ett bättre underlag vid dimensionering. Detta då högsta värmelasten under ett dygnskiljer sig från dygnsmedellasten. Att implementera modell med timdatat ökar känsligheten imodellen och ställer högre krav på den inhämtade kundlastdatat. Mätare i fjärrvärmecentralerbör ses över för säkerställning av god mätupplösning och mätprecision.

Fotboll och ekonomi är två ämnen som tillsammans utgör en stor del av världens intressen. Det är dessutom ämnen som under 2000-talet oftast blir nämnda i samma mening. Oftast talas det om att pengar inte kan köpa allt, och i fotbollsvärlden handlar det om att man inte kan köpa en ligatitel. Denna studie fokuserar på att undersöka hur ekonomisk ställning i ett svenskt elitfotbollslag kan påverka deras slutliga tabellplacering i Allsvenskan och Superettan. Inledningsvis diskuteras det om svensk fotbolls ställning i världen, tidigare studier och på vilket sätt denna studie tar sig ann ämnet. Relevant teori kring ämnet presenteras och även en förklaring kring urvalet av olika ekonomiska nyckeltal. Undersökningen omfattar totalt sex olika ekonomiska nyckeltal tagna från årsredovisningar från sex olika lag under är period på nio år. Dessa nyckeltal sätts i relation till lagens tabellplaceringar i en analys av beskrivande statistik, enkel linjär regression och multipel regression. Genom en analys av beskrivande statistik kring den insamlade datan visade studien på att en stark ekonomi kan ge en fördel i jakten på ligaguldet. Resultatet från den enkla linjära regressionen gav en analys på hur de olika nyckeltalen påverkar tabellplacering var för sig. Slutligen av resultaten från den multipla regressionen fick studien svar på vilka av de ekonomiska nyckeltalen som tillsammans bäst kan förklara vilken tabellplacering laget tar. För framtida studier föreslås en undersökning med ett större urval av variabler relaterade till ekonomisk styrka och prestation.

Cette thèse traite les problèmes statistiques suivants : la sélection de variables dans le modèle de régression linéaire en grande dimension, le clustering dans le modèle de mélange Gaussien, quelques effets de l'adaptabilité sous l'hypothèse de parcimonie ainsi que la simulation des processus Gaussiens.Sous l'hypothèse de parcimonie, la sélection de variables correspond au recouvrement du "petit" ensemble de variables significatives. Nous étudions les propriétés non-asymptotiques de ce problème dans la régression linéaire en grande dimension. De plus, nous caractérisons les conditions optimales nécessaires et suffisantes pour la sélection de variables dans ce modèle. Nous étudions également certains effets de l'adaptation sous la même hypothèse. Dans le modèle à vecteur parcimonieux, nous analysons les changements dans les taux d'estimation de certains des paramètres du modèle lorsque le niveau de bruit ou sa loi nominale sont inconnus.Le clustering est une tâche d'apprentissage statistique non supervisée visant à regrouper des observations proches les unes des autres dans un certain sens. Nous étudions le problème de la détection de communautés dans le modèle de mélange Gaussien à deux composantes, et caractérisons précisément la séparation optimale entre les groupes afin de les recouvrir de façon exacte. Nous fournissons également une procédure en temps polynomial permettant un recouvrement optimal des communautés.Les processus Gaussiens sont extrêmement utiles dans la pratique, par exemple lorsqu'il s'agit de modéliser les fluctuations de prix. Néanmoins, leur simulation n'est pas facile en général. Nous proposons et étudions un nouveau développement en série à taux optimal pour simuler une grande classe de processus Gaussiens. / This PhD thesis deals with the following statistical problems: Variable selection in high-Dimensional Linear Regression, Clustering in the Gaussian Mixture Model, Some effects of adaptivity under sparsity and Simulation of Gaussian processes.Under the sparsity assumption, variable selection corresponds to recovering the "small" set of significant variables. We study non-asymptotic properties of this problem in the high-dimensional linear regression. Moreover, we recover optimal necessary and sufficient conditions for variable selection in this model. We also study some effects of adaptation under sparsity. Namely, in the sparse vector model, we investigate, the changes in the estimation rates of some of the model parameters when the noise level or its nominal law are unknown.Clustering is a non-supervised machine learning task aiming to group observations that are close to each other in some sense. We study the problem of community detection in the Gaussian Mixture Model with two components, and characterize precisely the sharp separation between clusters in order to recover exactly the clusters. We also provide a fast polynomial time procedure achieving optimal recovery.Gaussian processes are extremely useful in practice, when it comes to model price fluctuations for instance. Nevertheless, their simulation is not easy in general. We propose and study a new rate-optimal series expansion to simulate a large class of Gaussian processes.

No description available.