Search

For multivariate regressors, integrating the Nadaraya–Watson regression smoother produces estimators of the lower-dimensional marginal components that are asymptotically normally distributed, at the optimal rate of convergence. Some heuristics, based on consistency of the pilot estimator, suggested that the estimator would not converge at the optimal rate of convergence in the presence of more than four covariates. This paper shows first that marginal integration with its internally normalized counterpart leads to rate-optimal estimators of the marginal components. We introduce the necessary modifications and give central limit theorems. Then, it is shown that the method apply also to more general models, in particular we discuss feasible estimation of partial linear models. The proofs reveal that the pilot estimator shall over-smooth the variables to be integrated, and, that the resulting estimator is itself a lower-dimensional regression smoother. Hence, finite sample properties of the estimator are comparable to those of low-dimensional nonparametric regression. Further advantages when starting with the internally normalized pilot estimator are its computational attractiveness and better performance (compared to its classical counterpart) when the covatiates are correlated and nonuniformly distributed. Simulation studies underline the excellent performance in comparison with so far known methods.

Consider an inverse problem with random noise where we want to estimate a function f. Moreover, suppose that the operator A that we need to invert is not completely known: we know its eigenfunctions and observe its singular values but with some noise. To construct our estimator θ, we minimize a modification of an unbiased risk estimator. We obtain some non-asymptotic exact oracle inequalities. Considering smooth functions in some standard classes of functions, we prove that θ is asymptotically minimax among a given class of estimators.

SummaryWhen modelling multivariate financial data, the problem of structural learning is compounded by the fact that the covariance structure changes with time. Previous work has focused on modelling those changes by using multivariate stochastic volatility models. We present an alternative to these models that focuses instead on the latent graphical structure that is related to the precision matrix. We develop a graphical model for sequences of Gaussian random vectors when changes in the underlying graph occur at random times, and a new block of data is created with the addition or deletion of an edge. We show how a Bayesian hierarchical model incorporates both the uncertainty about that graph and the time variation thereof.

Assessments of environmental and territorial justice are similar in that both assess whether empirical relations between the spatial arrangement of undesirable hazards (or desirable public goods and services) and sociodemographic groups are consistent with notions of social justice, evaluating the spatial distribution of benefits and burdens (outcome equity) and the process that produces observed differences (process equity). Using proximity to major highways in New York City as a case study, we review methodological issues pertinent to both fields and discuss choice and computation of exposure measures, but focus primarily on measures of inequity. We present inequity measures computed from the empirically estimated joint distribution of exposure and demographics and compare them with traditional measures such as linear regression, logistic regression, and Theil's entropy index. We find that measures computed from the full joint distribution provide more unified, transparent, and intuitive operational definitions of inequity and show how the approach can be used to structure siting and decommissioning decisions.

Despite improved control measures, Ebola remains a serious public health risk in African regions where recurrent outbreaks have been observed since the initial epidemic in 1976. Using epidemic modeling and data from two well-documented Ebola outbreaks (Congo 1995 and Uganda 2000), we estimate the number of secondary cases generated by an index case in the absence of control interventions R0. Our estimate of R0 is 1.83 (SD 0.06) for Congo (1995) and 1.34 (SD 0.03) for Uganda (2000). We model the course of the outbreaks via an SEIR (susceptible-exposed-infectious-removed) epidemic model that includes a smooth transition in the transmission rate after control interventions are put in place. We perform an uncertainty analysis of the basic reproductive number R0 to quantify its sensitivity to other disease-related parameters. We also analyse the sensitivity of the final epidemic size to the time interventions begin and provide a distribution for the final epidemic size. The control measures implemented during these two outbreaks (including education and contact tracing followed by quarantine) reduce the final epidemic size by a factor of 2 relative the final size with a 2-week delay in their implementation.

This article presents techniques for constructing classifiers that combine statistical information from training data with tangent approximations to known transformations; it demonstrates the techniques by applying them to a face recognition task. Our approach is to build Bayes classifiers with approximate class-conditional probability densities for measured data. The high dimension of the measurements in modern classification problems such as speech or image recognition makes inferring probability densities from feasibly sized training datasets difficult. We address the difficulty by imposing severely simplifying assumptions and exploiting a priori information about transformations to which classification should be invariant. For the face recognition task, we used a five-parameter group of such transformations consisting of rotation, shifts, and scalings. On the face recognition task, a classifier based on our techniques has an error rate that is 20% lower than that of the best algorithm in a reference s...

SummaryThe paper presents a general strategy for selecting the bandwidth of nonparametric regression estimators and specializes it to local linear regression smoothers. The procedure requires the sample to be divided into a training sample and a testing sample. Using the training sample we first compute a family of regression smoothers indexed by their bandwidths. Next we select the bandwidth by minimizing the empirical quadratic prediction error on the testing sample. The resulting bandwidth satisfies a finite sample oracle inequality which holds for all bounded regression functions. This permits asymptotically optimal estimation for nearly any regression function. The practical performance of the method is illustrated by a simulation study which shows good finite sample behaviour of our method compared with other bandwidth selection procedures.

Writers are often viewed as having an inherent style which can serve as a literary fingerprint. By quantifying relevant features related to literary style, one may hope to classify written works and even attribute authorship to newly discovered texts. Beyond its intrinsic interest, the study of literary styles presents the opportunity to introduce and motivate many standard multivariate statistical techniques. Today the statisti- cal analysis of literary styles is made much simpler by the wealth of real data readily available from the Internet. This paper presents an overview and brief history of the analysis of literary styles. In addition we use canonical discriminant analyis and prin- cipal component analysis to identify structure in the data and distinguish authorship.

International audience; Multivariate nonparametric smoothers are adversely impacted by the sparseness of data in higher dimension, also known as the curse of dimensionality. Adaptive smoothers, that can exploit the underlying smoothness of the regression function, may partially mitigate this effect. We present an iterative procedure based on traditional kernel smoothers, thin plate spline smoothers or Duchon spline smoother that can be used when the number of covariates is important. However the method is limited to small sample sizes (n < 2,000) and we will propose some thoughts to circumvent that problem using, for example,pre-clustering of the data. Applications considered here are image denoising.

This paper presents a practical and simple fully nonparametric multivariate smoothing procedure that adapts to the underlying smoothness of the true regression function. Our estimator is easily computed by successive application of existing base smoothers (without the need of selecting an optimal smoothing parameter), such as thin-plate spline or kernel smoothers. The resulting smoother has better out of sample predictive capabilities than the underlying base smoother, or competing structurally constrained models (GAM) for small dimension (3 < d < 8) and moderate sample size (n < 800). Moreover our estimator is still useful when (d> 10) and to our knowledge, no other adaptive fully nonparametric regression estimator is available without constrained assumption such as additivity for example. On a real example, the Boston Housing Data, our method reduces the out of sample prediction error by 20 %. An R package ibr, available at CRAN, implements the proposed multivariate nonparametric method in R., 26 pages, 6 figures

The authors consider the problem of estimating a regression function go involving several variables by the closest functional element of a prescribed class G that is closest to it in the L1 norm. They propose a new estimator ĝ based on independent observations and give explicit finite sample bounds for the L1distance between ĝg and go. They apply their estimation procedure to the problem of selecting the smoothing parameter in nonparametric regression.

We study the question of environmental equity via generalized linear modeling for the metropolitan New York City region and ask whether, after accounting for socioeconomic status, particular racial/ethnic populations bear a disproportionate burden of hosting environmentally undesirable sites. Our data consist of population demographics for 2216 census tracts linked to 354 environmentally undesirable facilities, including toxic release inventory sites, hazardous waste treatment, storage, and disposal facilities, and other common urban problem sites such as landfills, incinerators, bus garages and sewage treatment plants. Using generalized linear and additive modeling techniques, we find that racial/ethnic demographics, in particular the Hispanic percentage of a tract's population, are significantly associated with the presence of potentially environmentally adverse sites. This leads us to the conclusion that, over the whole metropolitan New York City area, the Hispanic population is proximate to more sites than other populations. At the same time, we find that both Hispanics and African-Americans are more proximate to these sites in the Bronx and Queens. However, we also find indications that Hispanics and African-Americans are less likely to be proximate to the sites in Manhattan. We establish an empirical relationship that warrants additional study in order to establish the causes for the population distribution and whether a basis for a claim of discrimination exists.

Protecting the Nation’s infrastructure from intentional attacks and natural disasters, including extreme weather events and climate change, is a major national security concern that has only become more critical since the terrorist attacks on September 11, 2001 (This chapter focuses on the work performed at LANL concerning the protection of the critical infrastructures of the United States (the ‘Nation’); however the modeling concepts discussed here are generally applicable). Understanding potential weaknesses of infrastructure assets and how interdependencies across critical infrastructure affect their behavior is essential to predicting and mitigating single and cascading failures, as well as to planning for response and recovery and future infrastructure development. Modeling and simulation (M&S) is an indispensable part of characterizing this complex system of systems and anticipating its response to disruptions. With the advent of more sophisticated infrastructure M&S capabilities, the possible applications have expanded to include the security challenges faced by the U.S. military, which relies on sustainable energy resources and needs to address environmental challenges and husband its water resources. Another key area where infrastructure modeling can play a critical role is in addressing global warming concerns given changes in available technology, evolution of the energy mix toward renewable resources, and many other infrastructure-related factors.

The Aquificales are thermophilic microorganisms that inhabit hydrothermal systems worldwide and are considered one of the earliest lineages of the domain Bacteria. We analyzed metagenome sequence obtained from six thermal “filamentous streamer” communities (∼40 Mbp per site), which targeted three different groups of Aquificales found in Yellowstone National Park (YNP). Unassembled metagenome sequence and PCR-amplified 16S rRNA gene libraries revealed that acidic, sulfidic sites were dominated by Hydrogenobaculum (Aquificaceae) populations, whereas the circum-neutral pH (6.5–7.8) sites containing dissolved sulfide were dominated by Sulfurihydrogenibium spp. (Hydrogenothermaceae). Thermocrinis (Aquificaceae) populations were found primarily in the circum-neutral sites with undetectable sulfide, and to a lesser extent in one sulfidic system at pH 8. Phylogenetic analysis of assembled sequence containing 16S rRNA genes as well as conserved protein-encoding genes revealed that the composition and function of these communities varied across geochemical conditions. Each Aquificales lineage contained genes for CO2 fixation by the reverse-TCA cycle, but only the Sulfurihydrogenibium populations perform citrate cleavage using ATP citrate lyase (Acl). The Aquificaceae populations use an alternative pathway catalyzed by two separate enzymes, citryl-CoA synthetase (Ccs), and citryl-CoA lyase (Ccl). All three Aquificales lineages contained evidence of aerobic respiration, albeit due to completely different types of heme Cu oxidases (subunit I) involved in oxygen reduction. The distribution of Aquificales populations and differences among functional genes involved in energy generation and electron transport is consistent with the hypothesis that geochemical parameters (e.g., pH, sulfide, H2, O2) have resulted in niche specialization among members of the Aquificales.

International audience; Multivariate nonparametric smoothers, such as kernel based smoothers and thin plate splines smoothers, are adversely impacted by the sparseness of data in high dimension, also known as the curse of dimensionality. Adaptive smoothers, that can exploit the underlying smoothness of the regression function, may partially mitigate this effect. This paper presents a comparative simulation study of a novel adaptive smoother (IBR) with competing multivariate smoothers available as package or function within the R language and environment for statistical computing. Comparison between the methods are made on simulated datasets of moderate size, from 50 to 200 observations, with two, five or 10 potential explanatory variables, and on a real dataset. The results show that the good asymptotic properties of IBR are complemented by a very good behavior on moderate sized datasets, results which are similar to those obtained with Duchon low rank splines.

We study the effects of randomness on competitions based on an elementary random process in which there is a finite probability that a weaker team upsets a stronger team. We apply this model to sports leagues and sports tournaments, and compare the theoretical results with empirical data. Our model shows that single-elimination tournaments are efficient but unfair: the number of games is proportional to the number of teams N, but the probability that the weakest team wins decays only algebraically with N. In contrast, leagues, where every team plays every other team, are fair but inefficient: the top $\sqrt{N}$ of teams remain in contention for the championship, while the probability that the weakest team becomes champion is exponentially small. We also propose a gradual elimination schedule that consists of a preliminary round and a championship round. Initially, teams play a small number of preliminary games, and subsequently, a few teams qualify for the championship round. This algorithm is fair and efficient: the best team wins with a high probability and the number of games scales as $N^{9/5}$, whereas traditional leagues require N^3 games to fairly determine a champion., Comment: 10 pages, 8 figures, reviews arXiv:physics/0512144, arXiv:physics/0608007, arXiv:cond-mat/0607694, arXiv:physics/0612217

Using polarized neutron reflectometry we measured the neutron spin-dependent reflectivity from four LaAlO3/SrTiO3 superlattices. Our results imply that the upper limit for the magnetization averaged over the lateral dimensions of the sample induced by an 11 T magnetic field at 1.7 K is less than 2 G. SQUID magnetometry of the neutron superlattice samples sporadically finds an enhanced moment, possibly due to experimental artifacts. These observations set important restrictions on theories which imply a strongly enhanced magnetism at the interface between LaAlO3 and SrTiO3.

Gamma detector counts are included in radiation portal monitors (RPM) to screen for illicit nuclear material. Gamma counts are sometimes smoothed to reduce variance in the estimated underlying true mean count rate, which is the “signal” in our context. Smoothing reduces total error variance in the estimated signal if the bias that smoothing introduces is more than offset by the variance reduction. An empirical RPM study for vehicle screening applications is presented for unsmoothed and smoothed gamma counts in low-resolution plastic scintillator detectors and in medium-resolution NaI detectors.

One of the standard methods of accounting for inter-population disease spread in equation-based epidemiology models is through a transportation operator. Implicit in the use of the transportation operator, however, is an assumption that daily travel volumes are small compared to overall population sizes, an assumption that can break down for modern rates of international travel or local commuter traffic. Alternative types of coupling have been proposed in the limit that trip durations are much shorter than the infectious period. We present an extension of these phenomenological models that relaxes both assumptions. We show that the approach produces more accurate results when assessing the impact of mitigative actions using modern travel volumes.

The 1990 Post-Enumeration Survey (PES) stratified the population into 1,392 subpopulations called post-strata based on location, race, tenure, sex and age, in the hope that these subpopulations were homogeneous in relation to factors affecting the Census coverage. Homogeneity is necessary to justify the use of the same adjustment factor for many, sometimes quite small, subgroups of the post-strata. With block-level data from the PES for sites around Detroit and Texas, we are able to examine empirically the extent to which this hope was realized. Using various measures, we find that between-block variation in erroneous enumeration and gross omission rates is about the same magnitude as, and largely in addition to, the corresponding between-post-stratum variation.

Within ray theory, travel times sample a finite set of points on the core-mantle boundary (CMB). Only if the CMB is smooth can we estimate its shape from travel times. The assumed degree of smoothness affects the uncertainty of the estimate. We find lower bounds on the uncertainty of the CMB shape by assuming (1) travel time errors are independent and identically distributed with zero mean and known variance, (2) linearization does not compromise accuracy, (3) event locations are exact, (4) radial velocity follows the IASP91 model, and (5) the only three-dimensional structure is a low-degree spherical harmonic perturbation to the CMB. We compute confidence intervals and simultaneous confidence envelopes for the CMB with a new technique based on reproducing kernels. Simultaneous confidence envelopes have a specified probability of containing the entire CMB, rather than just one point on the CMB, and thus allow inferences about the CMB shape. Suppose the standard deviation of travel time errors is 0.25 s (it is probably closer to 1.0 s), and that the CMB has a degree 4 spherical harmonic expansion. Then the width of a 95% confidence envelope around the least squares estimate of the CMB using 776,958 PKP phases and 84,947 PcP phases ranges from 0.32 km under Europe and Tonga to a maximum of 1.06 km under the mid-Pacific, with a median width of 0.58 km. If the CMB has structure up to degree 20, the envelope width ranges from 8.14 km to 245.0 km. It is wider than 40.5 km for 50% of the CMB, and wider than 67.3 km over 25% of the CMB. We construct a degree 20 CMB model extending almost 93 km into the lower mantle beneath the mid-Pacific that changes travel times by only 0.1 s rms.

Soft cosmic ray tomography has been shown to successfully discriminate materials with various density levels due to their ability to deeply penetrate matter, allowing sensitivity to atomic number, radiation length and density. Because the multiple muon scattering signal from high Z-materials is very strong, the technology is well suited to the detection of the illicit transportation of special and radiololgical nuclear materials. In addition, a recent detection technique based on measuring the lower energy particles that do not traverse the material (range radiography), allows to discriminate low and medium Z-materials. We have demonstrated it first using Monte Carlo simulations. More recently, using a Mini-Muon Tracker developed at Los Alamos National Laboratory, we performed various experiments to try out the radiation length technology. This paper presents the results from real experiments and evaluates the likelihood that soft cosmic ray tomography may be applied to detect high-explosives.

Random intersection graphs (RIGs) are an important random structure with applications in social networks, epidemic networks, blog readership, and wireless sensor networks. RIGs can be interpreted as a model for large randomly formed non-metric data sets. We analyze the component evolution in general RIGs, and give conditions on existence and uniqueness of the giant component. Our techniques generalize existing methods for analysis of component evolution: we analyze survival and extinction properties of a dependent, inhomogeneous Galton-Watson branching process on general RIGs. Our analysis relies on bounding the branching processes and inherits the fundamental concepts of the study of component evolution in Erd\H{o}s-R\'enyi graphs. The major challenge comes from the underlying structure of RIGs, which involves its both the set of nodes and the set of attributes, as well as the set of different probabilities among the nodes and attributes.

Charged particle tomography has been gaining ground due to its unique probing characteristics, and the ability of cosmic ray muons to deeply penetrate matter, allowing sensitivity to atomic number, radiation length and material density via detection of the probing particles before and after target traversal. But not all particles completely traverse the target volume. This paper explores the usefulness and information content in the stopped particles, referred to as soft particles. We show, in simulations, that the stopped particles are sensitive to low Z material and therefore provides complementary information to the scattering data.

We propose a universal approach for analysis and fast simulations of stiff stochastic biochemical networks, which rests on elimination of fast chemical species without a loss of information about mesoscopic, non-Poissonian fluctuations of the slow ones. Our approach is similar to the Born–Oppenheimer approximation in quantum mechanics and follows from the stochastic path integral representation of the cumulant generating function of reaction events. In applications with a small number of chemical reactions, it produces analytical expressions for cumulants of chemical fluxes between the slow variables. This allows for a low-dimensional, interpretable representation and can be used for high-accuracy, low-complexity coarse-grained numerical simulations. As an example, we derive the coarse-grained description for a chain of biochemical reactions and show that the coarse-grained and the microscopic simulations agree, but the former is 3 orders of magnitude faster.

The rising popularity of mobile devices, such as cellular phones and PDAs, has made them a lucrative playground for mobile malware propagation. One common infection vector exploited by these mobile malware is Bluetooth. In this paper, we propose an architecture called Blue-Watchdog that detects Bluetooth worm propagation in public areas based on statistical methods. To achieve fast and accurate Bluetooth worm detection, Blue-Watchdog monitors abrupt changes of average paging rate per Bluetooth device from both temporal and temporal-spatial perspectives. The temporal scheme relies on the CUSUM (Cumulative Sum) sequential test together with the generalized likelihood ratio (GLR), and the temporal-spatial scheme aims to identify spatial regions with abnormally frequent paging attempts. Experimental results show that Blue-Watchdog not only has low false alarm rates, but also effectively detects Bluetooth worms that spread quickly in areas where Bluetooth devices are greatly mixed due to high mobility and also those that propagate relatively slowly in a spatially constrained fashion.

This paper presents a a general framework for applying individual decision models to aggregated populations. Our approach is useful for modeling and predicting evacuation decisions from disasters, ranging from earthquakes, flooding and wild fires, to industrial emergencies like chemical spills or nuclear accidents, to reactions to terrorism attacks. The novelty of our approach is to apply well-documented household evacuation behavioral models to statistical accurate synthetic populations with detailed demographic information. Predictions of who evacuates and who does not evacuate, and the type and location of the selected shelters is useful for emergency management and planning. We illustrate it by applying our tools to predict emergency relocation behavior from hurricanes.

International audience; This paper introduces a computationally tractable density estimator that has the same asymptotic variance as the classical Nadaraya-Watson density estimator but whose asymptotic bias is zero. We achieve this result using a two stage estimator that applies a multiplicative bias correction to an oversmooth pilot estimator. Simulations show that our asymptotic results are available for samples as low as n = 50, where we see an improvement of as much as 20% over the traditionnal estimator. Mathematics Subject Classification. 62G07, 62G20.

We revisit the problem of computing the path with the minimum cost in terms of the expected number of link layer transmissions (including retransmissions) in wireless mesh networks. Unlike previous efforts, such as the popular ETX, we account for the fact that MAC protocols (including the IEEE 802.11 MAC) incorporate a finite number of transmission attempts per packet. This in turn leads to our key observation: the performance of a path depends not only on the number of the links on the path and the quality of its links, but also, on the relative positions of the links on the path. Based on this observation, we propose ETOP, a path metric that accurately captures the expected number of link layer transmissions required for reliable end-to-end packet delivery. We analytically compute ETOP, which is not trivial, since ETOP is a noncommutative function of the link success probabilities. Although ETOP is a more involved metric, we show that the problem of computing paths with the minimum ETOP cost can be solved by a greedy algorithm. We implement and evaluate a routing approach based on ETOP on a 25-node indoor mesh network. Our experiments show that the path selection with ETOP consistently results in superior TCP goodput (by over 50% in many cases) compared to path selection based on ETX. We also perform an in-depth analysis of the measurements to better understand why the paths selected by ETOP improve the TCP performance.

We measured the spin dependence of polarized neutrons reflected by a GaAs sample as it was optically pumped. This dependence was correlated with the helicity of the circularly polarized light and found to oscillate with neutron wave vector transfer. The data provide definitive evidence that optically induced nuclear polarization in GaAs is not uniform with depth. Quantitative analysis of the data shows that nuclear polarization is suppressed for tens of nanometers near the surface of GaAs.

In this paper, we present queue management algorithms with low implementation complexity that can partition the bandwidth of an outgoing link among flows in a high speed packet switch. These algorithms belong to a family of queue management schemes called sending rate estimate based queue management schemes (SREQM), which try to prevent congestion by effectively limiting flows based on their estimated sending rates. We also present techniques based on sampling and Bloom filters to further reduce the implementation overhead. The capability of the algorithms to protect responsive flows from non-responsive flows are confirmed by exhaustive analysis and simulations.

We revisit the problem of computing the path with the minimum cost in terms of the expected number of link layer retransmissions in wireless mesh networks. Unlike previous efforts (such as the popular ETX) we account for the fact that link layer protocols (such as the IEEE 802.11 MAC) incorporate a non-zero but finite number of retransmission attempts per packet. A key observation that motivates this work is that the performance of a path depends not only on the number of links on the path and their qualities, but also on the relative positions of the links on the path. In particular, the closer a lossy link to the destination, the higher is its impact on the performance of that path. We design a new path metricthat captures all of the above factors and we call this metric ETOP. In this paper, we provide a synopsis of the analytical computation of ETOP. We also implement a routing strategy based on ETOP on a 25-node experimental testbed and provide sample results to showcase the performance with ETOP.

League competition is investigated using random processes and scaling techniques. In our model, a weak team can upset a strong team with a fixed probability. Teams play an equal number of head-to-head matches and the team with the largest number of wins is declared to be the champion. The total number of games needed for the best team to win the championship with high certainty $T$ grows as the cube of the number of teams $N$, i.e., $T\ensuremath{\sim}{N}^{3}$. This number can be substantially reduced using preliminary rounds where teams play a small number of games and subsequently, only the top teams advance to the next round. When there are $k$ rounds, the total number of games needed for the best team to emerge as champion, ${T}_{k}$, scales as follows, ${T}_{k}\ensuremath{\sim}{N}^{{\ensuremath{\gamma}}_{k}}$ with ${\ensuremath{\gamma}}_{k}={[1\ensuremath{-}{(2∕3)}^{k+1}]}^{\ensuremath{-}1}$. For example, ${\ensuremath{\gamma}}_{k}=9∕5,27∕19,81∕65$ for $k=1,2,3$. These results suggest an algorithm for how to infer the best team using a schedule that is linear in $N$. We conclude that league format is an ineffective method of determining the best team, and that sequential elimination from the bottom up is fair and efficient.

At the outset of an influenza pandemic, early estimates of the number of secondary cases generated by a primary influenza case (reproduction number, R) and its associated uncertainty can help determine the intensity of interventions necessary for control. Using a compartmental model and hospital notification data of the first two waves of the Spanish flu pandemic in Geneva, Switzerland in 1918, we estimate the reproduction number from the early phase of the pandemic waves. For the spring and fall pandemic waves, we estimate reproduction numbers of 1:57 (95% CI: 1:45, 1:70) and 3:10 (2:81, 3:39), respectively, from the initial epidemic phase comprising the first 10 epidemic days of the corresponding wave. Estimates of the variance of our point estimates of R were computed via a parametric bootstrap. We compare these estimates with others obtained using different observation windows to provide insight into how early into an epidemic the reproduction number can be estimated.

We propose a rigorous mathematical interpretation of a novel family of active queue management schemes, called sending rate estimate based queue management (SREQM) scheme, that aims to provide fair bandwidth allocation to all the flows in a router by estimating the flow sending rates, while maintaining only minimal per-flow state information. We propose an optimized implementation of SREQM, called fair sending rate estimate based queue management (FSREQM) scheme, and show through comparative simulation that FRESQM is the only scheme among those tested that successfully prevents bandwidth abuse while maintaining high link utilization.

The alarm that worms start to spread on increasingly popular mobile devices calls for an in-depth investigation of their propagation dynamics. In this paper, we study how mobility patterns affect Bluetooth worm spreading speeds. We find that the impact of mobility patterns is substantial over a large set of of changing Bluetooth and worm parameters. For instance, a mobility model under which devices move among a fixed set of activity locations can result in worm propagation speeds four times faster than a classical mobility model such as the random walk model. Our investigation reveals that the key factors affecting Bluetooth worm propagation speeds include spatial distributions of nodes, link duration distributions, degrees to which devices are mixed together, and even the burstiness of successive links.

We propose ESREQM (Efficient Sending Rate Estimation Queue Management), a novel active queue management scheme that achieves almost perfect max-min fairness among flows with minimum (constant) per-flow state and a constant number of CPU operations to handle an incoming packet using a single queue with very low buffer requirements. ESREQM estimates sending rates of flows through a history discounting process that allows it to guarantee max-min fairness by automatically adapting parameters. It can also be used to punish non-responsive flows. The per-flow state is limited to a single value per flow, which allows the flow memory to be in SRAM, thereby making packet processing scalable with link speeds. ESREQM results in good link utilization with low buffer size requirements because it provably desynchronizes TCP flows as a by-product. We show our results through a mixture of analysis and simulation. Our scheme does not make assumptions on what transport protocols are used.

Background Syndromic surveillance (SS) can potentially contribute to outbreak detection capability by providing timely, novel data sources. One SS challenge is that some syndrome counts vary with season in a manner that is not identical from year to year. Our goal is to evaluate the impact of inconsistent seasonal effects on performance assessments (false and true positive rates) in the context of detecting anomalous counts in data that exhibit seasonal variation. Methods To evaluate the impact of inconsistent seasonal effects, we injected synthetic outbreaks into real data and into data simulated from each of two models fit to the same real data. Using real respiratory syndrome counts collected in an emergency department from 2/1/94–5/31/03, we varied the length of training data from one to eight years, applied a sequential test to the forecast errors arising from each of eight forecasting methods, and evaluated their detection probabilities (DP) on the basis of 1000 injected synthetic outbreaks. We did the same for each of two corresponding simulated data sets. The less realistic, nonhierarchical model's simulated data set assumed that "one season fits all," meaning that each year's seasonal peak has the same onset, duration, and magnitude. The more realistic simulated data set used a hierarchical model to capture violation of the "one season fits all" assumption. Results This experiment demonstrated optimistic bias in DP estimates for some of the methods when data simulated from the nonhierarchical model was used for DP estimation, thus suggesting that at least for some real data sets and methods, it is not adequate to assume that "one season fits all." Conclusion For the data we analyze, the "one season fits all " assumption is violated, and DP performance claims based on simulated data that assume "one season fits all," for the forecast methods considered, except for moving average methods, tend to be optimistic. Moving average methods based on relatively short amounts of training data are competitive on all three data sets, but are particularly competitive on the real data and on data from the hierarchical model, which are the two data sets that violate the "one season fits all" assumption.

The 1918 influenza pandemic known as the “Spanish Flu” has been the worst in recent history with estimated worldwide mortality ranging from 20 to 100 million deaths. Using epidemic modeling and hospital notification data during the 1918 influenza pandemic in the Canton of Geneva, Switzerland, we estimated the reproductive numbers of the first and second waves of influenza infection to be R1 = 1.49 (95% CI: 1.45–1.53) and R2 = 3.75 (95% CI: 3.57–3.93), respectively. Our estimates indicate that containment of the next influenza pandemic could require strict interventions that include effective isolation strategies in hospitals and reductions in the susceptibility of the general population.

The paths of muons traveling through matter are affected by Coulomb multiple scattering. The magnitude of that interaction depends on the radiation length of the traversed medium, with high-Z matter scattering more than low-Z matter. The net effect on the path of a muon through an object is both a change in the direction and an orthogonal displacement orthogonal. Both these quantities can be measured experimentally for individual muons, opening the possibility of tomographic reconstruction of the radiation lengths of unknown complex objects. We present a framework to characterize the ability to reconstruct the scattering density from muon scattering data for detecting and locating areas of high Z material. Our analysis shows the importance of having detectors with large aperture, and the importance of using both the change in angle and displacement

We have built a detector capable of locating high Z objects in the sampling (middle) region of the detector. As atomic number increases, radiation length rapidly decreases, yielding larger variance in scattering angle. Cosmic ray muon tomography works by tracking muons above the sampling region, and tracking them below the region as well. The difference between the two trajectories yield information, via the muon scattering variance, of the materials contained within the sampling region [Borozdin, K, et al., 2003]. One of most important aspects of cosmic ray tomography is minimizing exposure time. The cosmic ray flux is about 1 cm-2 min-1, and the goal is to use them for detecting high-density materials as quickly as possible. This involves using all of the information possible to reconstruct tracks with redundant detectors. Detector scattering residuals yield a low precision measurement of muon energy. Knowing the rough energy of an incoming particle will yield more precisely the expected scattering variance (currently the expectation value of ~3 GeV is used).

In many applications the time available for nuclear materials detection is short, the signals are weak, and the backgrounds are large, variable, and difficult to control. We develop optimal strategies for nuclear search in these difficult conditions, and demonstrate these strategies in experiments. Model-driven algorithms promise to reduce search time by order of magnitude, with increased reliability of detection and reduced number of false positives. These strategies can be applied to various nuclear search scenarios, for both mobile and stationary detectors, hand-held detectors and sensors on robotic platforms. Our methods can be used against nuclear smugglers and terrorists, for safeguards and non-proliferation treaty monitoring, as well as in other situations where radioactive sources need to be found.

We present a technique for combining prior knowledge about transformations that should be ignored with a covariance matrix estimated from training data to make an improved Mahalanobis distance classifier. Modern classification problems often involve objects represented by high-dimensional vectors or images (for example, sampled speech or human faces). The complex statistical structure of these representations is often difficult to infer from the relatively limited training data sets that are available in practice. Thus, we wish to efficiently utilize any available a priori information, such as transformations or the representations with respect to which the associated objects are known to retain the same classification (for example, spatial shifts of an image of a handwritten digit do not alter the identity of the digit). These transformations, which are often relatively simple in the space of the underlying objects, are usually nonlinear in the space of the object representation, making their inclusion within the framework of a standard statistical classifier difficult. Motivated by prior work of Simard et al. (1998; 2000), we have constructed a new classifier which combines statistical information from training data and linear approximations to known invariance transformations. When tested on a face recognition task, performance was found to exceed by a significant margin that of the best algorithm in a reference software distribution.

The asymptotic risk of efficient estimators with Kullback–Leibler loss in smoothly parametrized statistical models is $k/2_n$, where $k$ is the parameter dimension and $n$ is the sample size. Under fairly general conditions, we given a simple information-theoretic proof that the set of parameter values where any arbitrary estimator is superefficient is negligible. The proof is based on a result of Rissanen that codes have asymptotic redundancy not smaller than $(k/2)\log n$, except in a set of measure 0.

Let $X_1, X_2, \dots, X_n$ be an i.i.d. sample from the Poisson mixture distribution $p(x) = (1/x!) \int_0^{\infty} s^x e^{-s}f(s) ds$. Rates of convergence in mean integrated squared error (MISE) of orthogonal series estimators for the mixing density f supported on $[a, b]$ are studied. For the Hölder class of densities whose rth derivative is Lipschitz $\alpha$, the MISE converges at the rate $(\log n/ \log \log n)^{-2(r +\alpha)}$. For Sobolev classes of densities whose rth derivative is square integrable, the MISE converges at the rate $(\log n/ \log \log n)^{-2r}$. The estimator is adaptive over both these classes. ¶ For the Sobolev class, a lower bound on the minimax rate of convergence is $(\log n/ \log \log n)^{-2r}$, and so the orthogonal polynomial estimator is rate optimal.

In most treatments of nonparametric regression, it is assumed that the marginal density of the explanatory variables is strictly bounded away from zero and infinity. This note investigates the pointwise asymptotics for nonparametric regression when this assumption fails, that is, the marginal density of the explanatory variable has either an isolated zero or a pole at the point of interest.

A conservative finite-sample simultaneous confidence envelope for a density can be found by solving a finite set of finite-dimensional linear programming problems if the density is known to be monotonic or to have at most $k$ modes relative to a positive weight function. The dimension of the problems is at most $(n/\log n)^{1/3}$, where $n$ is the number of observations. The linear programs find densities attaining the largest and smallest values at a point among cumulative distribution functions in a confidence set defined using the assumed shape restriction and differences between the empirical cumulative distribution function evaluated at a subset of the observed points. Bounds at any finite set of points can be extrapolated conservatively using the shape restriction. The optima are attained by densities piecewise proportional to the weight function with discontinuities at a subset of the observations and at most five other points. If the weight function is constant and the density satisfies a local Lipschitz condition with exponent $\varrho$, the width of the bounds converges to zero at the optimal rate $(\log n/n)^{\varrho/(1+2\varrho)}$ outside every neighborhood of the set of modes, if a "bandwidth" parameter is chosen correctly. The integrated width of the bounds converges at the same rate on intervals where the density satisfies a Lipschitz condition if the intervals are strictly within the support of the density. The approach also gives algorithms to compute confidence intervals for the support of monotonic densities and for the mode of unimodal densities, lower confidence intervals on the number of modes of a distribution and conservative tests of the hypothesis of $k$-modality. We use the method to compute confidence bounds for the probability density of aftershocks of the 1984 Morgan Hill, CA, earthquake, assuming aftershock times are an inhomogeneous Poisson point process with decreasing intensity.