Master of Science / Department of Statistics / Trevor Hefley / Binary regression models for spatial data are commonly used in disciplines such as epidemiology and ecology. Many spatially-referenced binary data sets suffer from location error, which occurs when the recorded location of an observation differs from its true location. When location error occurs, values of the covariates associated with the true spatial locations of the observations cannot be obtained. We show how a change of support (COS) can be applied to regression models for binary data to provide bias-corrected coefficient estimates when the true values of the covariates are unavailable, but the unknown location of the observations are contained within non-overlapping polygons of any geometry. The COS accommodates spatial and non-spatial covariates and preserves the convenient interpretation of methods such as logistic and probit regression. Using a simulation experiment, we compare binary regression models with a COS to naive approaches that ignore location error. We illustrate the flexibility of the COS by modeling individual-level disease risk in a population using a binary data set where the location of the observations are unknown, but contained within administrative units. Our simulation experiment and data illustration corroborate that conventional regression models for binary data which ignore location error are unreliable, but that the COS can be used to eliminate bias while preserving model choice.

Vehicular communication is the enabling technology for the development of the intelligent transportation systems (ITS), which aims to improve the efficiency and safety of transportation. It can be used for a variety of useful applications such as adaptive traffic control, coordinated braking, emergency messaging, peer-to-peer networking for infotainment services and automatic toll collection etc... Accurate yet simple models for vehicular networks are required in order to understand and optimize their operation. For reliable communication between vehicles, the spectrum access is coordinated via carrier sense multiple access (CSMA) protocol. Existing models either use a simplified network abstraction and access control scheme for analysis or depend on simulation studies. Therefore it is important to develop an analytical model for CSMA coordinated communication between vehicles. In the first part of the thesis, stochastic geometry is exploited to develop a modeling framework for CSMA coordinated inter-vehicle communication (IVC) in a multi-lane highway scenario. The performance of IVC is studied in multi-lane highways taking into account the inter-lane separations and the number of traffic lanes and it is shown that for wide multi-lane highways, the line abstraction model that is widely used in literature loses accuracy and hence the analysis is not reliable. Since the analysis of CSMA in the vehicular setting makes the analysis intractable, an aggressive interference approximation and a conservative interference approximation is proposed for the probability of transmission success. These approximations are tight in the low traffic and high traffic densities respectively. In the subsequent part of the thesis, the developed model is extended to multi-hop IVC because several vehicular applications require going beyond the local communication and efficiently disseminate information across the roads via multi-hops. Two well-known greedy packet forwarding schemes are studied, that impose different tradeoffs between per-hop transmission success probability and forward packet progress, namely, the most forward with fixed radius (MFR) and the nearest with forward progress (NFP). In particular, a tractable and accurate modeling framework is developed to characterize the per-hop transmission success probability and the average forward progress for vehicular networks in a multi-lane highway setup. The developed model reveals the interplay between the spectrum sensing threshold of the CSMA protocol and the packet forwarding scheme. A new performance metric is defined, denoted as the aggregate packet progress (APP), which is a dimensionless quantity that captures the tradeoffs between the spatial frequency reuses efficiency, the per-hop transmission success probability, and the per-hop forward progress of the packets. To this end, in contrary to existing studies, the results show that with the proper manipulation of CSMA threshold, the MFR achieves the highest APP.

Exact characterization and performance analysis of wireless networks should incorporate dependencies or correlations in space and time, i.e., study how the network performance varies spatially and temporally while having prior information about the performance at previous locations and time slots. This spatio-temporal correlation in wireless networks is usually characterized by studying metrics such as joint coverage probability at two spatial locations/time slots or spatio-temporal correlation coefficient. While developing models and analytical expressions for studying the two extreme cases of spatio-temoral correlation - i) uncorrelated scenario and ii) fully correlated scenario are easier, studying the intermediate case is non-trivial. In this thesis, we develop realistic and tractable analytical frameworks based on random spatial models (using tools from stochastic geometry) for modeling and analysis of correlation in cellular networks. With an ever increasing data demand, caching popular content in the storage of small cells (small cell caching) or the memory of user devices (device caching) is seen as a good alternative to offload demand from macro base stations and reduce backhaul loads. After providing generic results for traditional cellular networks, we study two applications exploiting spatio-temporal correlation in cache-enabled cellular networks. First, we determine the optimal cache content to be stored in the cache of a small cell network that maximizes the hit probability and minimizes the reception energy for the two extreme cases of correlation. Our results concretely demonstrate that the optimal cache contents are significantly different for the two correlation scenarios, thereby indicating the need of correlation-aware caching strategies. Second, we look at a distributed caching scenario in user devices and show that spatio-temporal correlation (user mobility) can be exploited to improve the network performance (in terms of coverage probability and local delay) significantly. / Master of Science

The availability of location estimates of mobile devices (MDs) is vital for several important applications such as law enforcement, disaster management, battlefield operations, vehicular communication, traffic safety, emergency response, and preemption. While global positioning system (GPS) is usually sufficient in outdoor clear sky conditions, its functionality is limited in urban canyons and indoor locations due to the absence of clear line-of-sight between the MD to be localized and a sufficient number of navigation satellites. In such scenarios, the ubiquitous nature of cellular networks makes them a natural choice for localization of MDs. Traditionally, localization in cellular networks has been studied using system level simulations by fixing base station (BS) geometries. However, with the increasing irregularity of the BS locations (especially due to capacity-driven small cell deployments), the system insights obtained by considering simple BS geometries may not carry over to real-world deployments. This necessitates the need to study localization performance under statistical (random) spatial models, which is the main theme of this work. In this thesis, we use powerful tools from stochastic geometry and point process theory to develop a tractable analytical model to study the localizability (ability to get a location fix) of an MD in single-tier and heterogeneous cellular networks (HetNets). More importantly, we study how availability of information about the location of proximate BSs at the MD impacts localizability. To this end, we derive tractable expressions, bounds, and approximations for the localizability probability of an MD. These expressions depend on several key system parameters, and can be used to infer valuable system insights. Using these expressions, we quantify the gains achieved in localizability of an MD when information about the location of proximate BSs is incorporated in the model. As expected, our results demonstrate that localizability improves with the increase in density of BS deployments. / Master of Science / Location based services form an integral part of vital day-to-day applications such as traffic control, emergency response, and navigation. Traditionally, users have relied on the global positioning system system (GPS) for localizing a device. GPS systems rely on the availability of clear line-of-sight between the devices to be localized and a sufficient number of navigation satellites. Since it is not possible to have these line-of-sight links, especially in urban canyons and indoor locations, the ubiquity of cellular networks makes them a natural choice for localization. Typically, localization using cellular networks is studied using simulations, which are carried out by fixing the network configuration including the geometry of the base stations (BSs) as well as the number of BSs that participate in localization. This limits the scope of the results obtained since a change in the network configuration would mean that one must do another set of time consuming simulations with the new network parameters. This motivates the need to develop an analytical model to study the impact of fundamental system-design factors such as BS geometries, number of participating BSs, propagation effects, and channel conditions on localization in cellular networks. Such analysis would make it convenient to infer how changing these system parameters affects localization. In this thesis, we develop a general analytical model to study the localizability (ability of get a location fix) of a device in a cellular network. In particular, we study how information about the location of BSs in the proximity of the device to be localized affects localizability. We derive expressions for metrics such as the localizability probability of a device. Our results help quantify the gains achieved in localizability performance when information about the location of BSs in the vicinity of the device to be localized is available at the device. Our results concretely demonstrate that including this additional information significantly improves the localizability performance, especially in regions with dense BS deployments.

CoordenaÃÃo de AperfeiÃoamento de Pessoal de NÃvel Superior / Atualmente, a tecnologia celular està presente em todos os aspectos da vida cotidiana: lares, escritÃrios, indÃstrias, etc. Tal tecnologia teve um rÃpido crescimento durante as duas Ãltimas dÃcadas tentando acompanhar o aumento do volume de trÃfego nas redes de comunicaÃÃo sem-fio. Naturalmente, ao propor modelos mais realistas possÃveis, com o propÃsito de caracterizar fenÃmenos que afetam a qualidade do sinal ou o desempenho do sistema, novas ideias, concepÃÃes e outras ferramentas surgem para descrever tais situaÃÃes. Este à o caso da Geometria EstocÃstica ou, particularmente, o processo pontual de Poisson, o qual vem sendo frequentemente utilizado como um modelo de rede celular, a partir da localizaÃÃo aleatÃria dos nÃs na rede. Diante desta ferramenta matemÃtica, à possÃvel implantar estaÃÃes rÃdio base na rede externa celular, bem como pontos de acesso baseados em picocÃlulas, femtocÃlulas, etc. AlÃm disso, permite-se quantificar a interferÃncia, Ãrea de cobertura, probabilidade de outage, dentre outros. Estes resultados tambÃm levam em consideraÃÃo o impacto de mobilidade no desempenho de tais redes. Nesse contexto, este trabalho analisarà redes ad-hoc sem-fio propondo expressÃes analÃticas para as seguintes mÃtricas de caracterizaÃÃo de desempenho: interferÃncia e conectividade de transmissÃo. Essas mÃtricas levam em consideraÃÃo tanto a razÃo sinal-ruÃdo mais interferÃncia (signal-to-interference-plus-noise ratio (SINR)) como a razÃo sinal-interferÃnca (signal-to-interference ratio (SIR)), em que neste caso, a potÃncia de ruÃdo à considerada nula. Especificamente, o fenÃmeno interferÃncia serà caracterizado via modelo shot-noise segundo um processo pontual chamado de processo pontual marcado (marked point process (MPP)), sendo este mais realista do que o tradicional modelo de Poisson. AlÃm disso, este tipo de modelo incorpora os efeitos de propagaÃÃo de rÃdio de pequena e larga escala e sobretudo as diferentes tecnologias de detecÃÃo e tratamento de sinal. Paralelamente, adotaremos um canal de rÃdio com desvanecimento Nakagami-m. Por fim, o tratamento matemÃtico para o modelo proposto torna-se um fator desafiador deste trabalho, visto que, tais resultados generalizam alguns jà publicados na literatura, os quais adotam alguns parÃmetros menos realistas. / Currently, cellular technology is present in all aspects of everyday life: homes, offices, industries, etc. Such technology had grown rapidly over the last two decades trying to follow up with the increased traffic volume on the networks of wireless communication. Naturally, to propose possible more realistic models, with the purpose of characterizing phenomena that affect the signal quality or performance system, new ideas, concepts and other tools to describe such situations arise. This is the case of Stochastic Geometry or, particularly, the point process Poisson, which has been often used as a model for cellular network from the random node locations in the network. Faced with this mathematical tool, it is possible deploy base stations in cellular external network and access points based picocells, femtocells, etc. Moreover, it allows to quantify the interference, coverage area, outage probability, among others. These results also consider the impact of mobility on the performance of such networks. In this context, this thesis will analyze ad-hoc wireless networks offering analytical expressions for the following metrics of performance characterization: interference and transmission connections. These metrics take into account both signal-to-interference-plus-noise ratio (SINR) and signal-to-interference ratio (SIR), in which case, the noise power is considered null. Specifically, the interference phenomena will be characterized via shot-noise model according to a point process called marked point process (MPP), this being more realistic than the traditional Poisson model. Furthermore, this type of model incorporates effects of radio propagation small and large scale, mainly the different technologies for the detection and signal processing. In parallel, we will adopt a radio channel with Nakagami-m fading. Finally, the mathematical treatment for the proposed model becomes a challenging factor in this work, since such results generalize some already published in the literature, which adopt some less realistic parameters.

The aggregate interference distribution in cognitive radio networks is studied in a rigorous and analytical way using the popular Poisson point process model. While a number of results are available for this model for non-cognitive radio networks, cognitive radio networks present extra levels of difficulties for the analysis, mainly due to the exclusion region around the primary receiver, which are typically addressed via various ad-hoc approximations (e.g., based on the interference cumulants) or via the large-deviation analysis. Unlike the previous studies, we do not use here ad-hoc approximations but rather obtain the asymptotic interference distribution in a systematic and rigorous way, which also has a guaranteed level of accuracy at the distribution tail. This is in contrast to the large deviation analysis, which provides only the (exponential) order of scaling but not the outage probability itself. Unlike the cumulant-based analysis, our approach provides a guaranteed level of accuracy at the distribution tail. Additionally, our analysis provides a number of novel insights. In particular, we demonstrate that there is a critical transition point below which the outage probability decays only polynomially but above which it decays super-exponentially. This provides a solid analytical foundation to the earlier empirical observations in the literature and also reveals what are the typical ways outage events occur in different regimes. The analysis is further extended to include interference cancelation and fading (from a broad class of distributions). The outage probability is shown to scale down exponentially in the number of canceled nearest interferers in the below-critical region and does not change significantly in the above-critical one. The proposed asymptotic expressions are shown to be accurate in the non-asymptotic regimes as well.

The aggregate interference distribution in cognitive radio networks is studied in a rigorous and analytical way using the popular Poisson point process model. While a number of results are available for this model for non-cognitive radio networks, cognitive radio networks present extra levels of difficulties for the analysis, mainly due to the exclusion region around the primary receiver, which are typically addressed via various ad-hoc approximations (e.g., based on the interference cumulants) or via the large-deviation analysis. Unlike the previous studies, we do not use here ad-hoc approximations but rather obtain the asymptotic interference distribution in a systematic and rigorous way, which also has a guaranteed level of accuracy at the distribution tail. This is in contrast to the large deviation analysis, which provides only the (exponential) order of scaling but not the outage probability itself. Unlike the cumulant-based analysis, our approach provides a guaranteed level of accuracy at the distribution tail. Additionally, our analysis provides a number of novel insights. In particular, we demonstrate that there is a critical transition point below which the outage probability decays only polynomially but above which it decays super-exponentially. This provides a solid analytical foundation to the earlier empirical observations in the literature and also reveals what are the typical ways outage events occur in different regimes. The analysis is further extended to include interference cancelation and fading (from a broad class of distributions). The outage probability is shown to scale down exponentially in the number of canceled nearest interferers in the below-critical region and does not change significantly in the above-critical one. The proposed asymptotic expressions are shown to be accurate in the non-asymptotic regimes as well.

The purpose of the research presented here is twofold. The first component explores the probabilistic interpretation of Stein’s method, as introduced in Barbour (1988). This is done in the setting of random variable approximations. This probabilistic method, where the Stein equation is interpreted in terms of the generator of an underlying birth and death process having equilibrium distribution equal to that of the approximant, provides a natural explanation of why Stein’s method works. An open problem has been to use this generator approach to obtain bounds on the differences of the solution to the Stein equation. Uniform bounds on these differences produce Stein “magic” factors, which control the bounds. With the choice of unit per capita death rate for the birth and death process, we are able to produce a result giving a new Stein factor bound, which applies to a selection of distributions. The proof is via a probabilistic approach, and we also include a probabilistic proof of a Stein factor bound from Barbour, Holst and Janson (1992). These results generalise the work of Xia (1999), which applies to the Poisson distribution with unit per capita death rate. (For complete abstract open document)

Data traffic demand has been increasing exponentially and this trend will continue over theforeseeable future. This has forced operators to upgrade and densify their mobile networks toenhance their capacity. Future networks will be characterized by a dense deployment of different kinds of base stations (BSs) in a hierarchical cellular structure. However network densification requires extensive capital and operational investment which limits operator revenues and raises ecological concerns over greenhouse gas emissions. Although networks are planned to support peak traffic, traffic demand is actually highly variable in both space and time which makes it necessary to adapt network energy consumption to inevitable variations in traffic demand. In this thesis, stochastic geometry tools are used to perform simple and tractable analysis of thecoverage, rate and energy performance of homogeneous networks and heterogeneous networks(HetNets). BSs in each tier are located according to independent Poisson Point Processes(PPPs) to generate irregular topologies that fairly resemble practical deployment topologies. The homogeneous network is optimized to determine the optimal BS density and transmit power configuration that minimizes its area power consumption (APC) subject to both coverage and average rate constraints. Results show that optimal transmit power only depends on the BSpower consumption parameters and can be predetermined. Furthermore, various sleep modemechanisms are applied to the homogeneous network to adapt its APC to changes in userdensity. A centralized strategic scheme which prioritize BSs with the least number of usersenhances energy efficiency (EE) of the network. Due to the complexity of such a centralizedscheme, a distributed scheme which implements the strategic algorithm within clusters of BSsis proposed and its performance closely matches that of its centralized counterpart. It is more challenging to model the optimal deployment configuration per tier in a multi-tier HetNet. Appropriate assumptions are used to determine tight approximations of these deployment configurations that minimize the APC of biased and unbiased HetNets subject tocoverage and rate constraints. The optimization is performed for three different user associationschemes. Similar to the homogeneous network, optimal transmit power per tier also depends onBS power consumption parameters only and can also be predetermined. Analysis of the effect of biasing on HetNet performance shows appropriate biasing can further reduce the deploymentconfiguration (and consequently the APC) compared to an unbiased HetNet. In addition, biasing can be used to offload traffic from congesting and high-power macro BSs to low-power small BSs. If idle BSs are put into sleep mode, more energy is saved and HetNet EE improves. Moreover, appropriate biasing also enhances the EE of the HetNet.

The aggregate interference distribution in cognitive radio networks is studied in a rigorous and analytical way using the popular Poisson point process model. While a number of results are available for this model for non-cognitive radio networks, cognitive radio networks present extra levels of difficulties for the analysis, mainly due to the exclusion region around the primary receiver, which are typically addressed via various ad-hoc approximations (e.g., based on the interference cumulants) or via the large-deviation analysis. Unlike the previous studies, we do not use here ad-hoc approximations but rather obtain the asymptotic interference distribution in a systematic and rigorous way, which also has a guaranteed level of accuracy at the distribution tail. This is in contrast to the large deviation analysis, which provides only the (exponential) order of scaling but not the outage probability itself. Unlike the cumulant-based analysis, our approach provides a guaranteed level of accuracy at the distribution tail. Additionally, our analysis provides a number of novel insights. In particular, we demonstrate that there is a critical transition point below which the outage probability decays only polynomially but above which it decays super-exponentially. This provides a solid analytical foundation to the earlier empirical observations in the literature and also reveals what are the typical ways outage events occur in different regimes. The analysis is further extended to include interference cancelation and fading (from a broad class of distributions). The outage probability is shown to scale down exponentially in the number of canceled nearest interferers in the below-critical region and does not change significantly in the above-critical one. The proposed asymptotic expressions are shown to be accurate in the non-asymptotic regimes as well.