Showing papers by "University of Wisconsin-Madison published in 2006"

Model selection and estimation in regression with grouped variables

Abstract: Summary. We consider the problem of selecting grouped variables (factors) for accurate prediction in regression. Such a problem arises naturally in many practical situations with the multifactor analysis-of-variance problem as the most important and well-known example. Instead of selecting factors by stepwise backward elimination, we focus on the accuracy of estimation and consider extensions of the lasso, the LARS algorithm and the non-negative garrotte for factor selection. The lasso, the LARS algorithm and the non-negative garrotte are recently proposed regression methods that can be used to select individual variables. We study and propose efficient algorithms for the extensions of these methods for factor selection and show that these extensions give superior performance to the traditional stepwise backward elimination method in factor selection problems. We study the similarities and the differences between these methods. Simulations and real examples are used to illustrate the methods.

Paclitaxel-carboplatin alone or with bevacizumab for non-small-cell lung cancer

Abstract: bevacizumab, as compared with 10.3 months in the chemotherapy-alone group (hazard ratio for death, 0.79; P = 0.003). The median progression-free survival in the two groups was 6.2 and 4.5 months, respectively (hazard ratio for disease progression, 0.66; P<0.001), with corresponding response rates of 35% and 15% (P<0.001). Rates of clinically significant bleeding were 4.4% and 0.7%, respectively (P<0.001). There were 15 treatment-related deaths in the chemotherapy-plus-bevacizumab group, including 5 from pulmonary hemorrhage. Conclusions The addition of bevacizumab to paclitaxel plus carboplatin in the treatment of selected patients with non–small-cell lung cancer has a significant survival benefit with the risk of increased treatment-related deaths. (ClinicalTrials.gov number, NCT00021060.)

The relationship between Precision-Recall and ROC curves

Abstract: Receiver Operator Characteristic (ROC) curves are commonly used to present results for binary decision problems in machine learning. However, when dealing with highly skewed datasets, Precision-Recall (PR) curves give a more informative picture of an algorithm's performance. We show that a deep connection exists between ROC space and PR space, such that a curve dominates in ROC space if and only if it dominates in PR space. A corollary is the notion of an achievable PR curve, which has properties much like the convex hull in ROC space; we show an efficient algorithm for computing this curve. Finally, we also note differences in the two types of curves are significant for algorithm design. For example, in PR space it is incorrect to linearly interpolate between points. Furthermore, algorithms that optimize the area under the ROC curve are not guaranteed to optimize the area under the PR curve.

Radiotherapy plus Cetuximab for Squamous-Cell Carcinoma of the Head and Neck

Abstract: BACKGROUND We conducted a multinational, randomized study to compare radiotherapy alone with radiotherapy plus cetuximab, a monoclonal antibody against the epidermal growth factor receptor, in the treatment of locoregionally advanced squamous-cell carcinoma of the head and neck. METHODS Patients with locoregionally advanced head and neck cancer were randomly assigned to treatment with high-dose radiotherapy alone (213 patients) or high-dose radiotherapy plus weekly cetuximab (211 patients) at an initial dose of 400 mg per square meter of body-surface area, followed by 250 mg per square meter weekly for the duration of radiotherapy. The primary end point was the duration of control of locoregional disease; secondary end points were overall survival, progression-free survival, the response rate, and safety. RESULTS The median duration of locoregional control was 24.4 months among patients treated with cetuximab plus radiotherapy and 14.9 months among those given radiotherapy alone (hazard ratio for locoregional progression or death, 0.68; P = 0.005). With a median follow-up of 54.0 months, the median duration of overall survival was 49.0 months among patients treated with combined therapy and 29.3 months among those treated with radiotherapy alone (hazard ratio for death, 0.74; P = 0.03). Radiotherapy plus cetuximab significantly prolonged progression-free survival (hazard ratio for disease progression or death, 0.70; P = 0.006). With the exception of acneiform rash and infusion reactions, the incidence of grade 3 or greater toxic effects, including mucositis, did not differ significantly between the two groups. CONCLUSIONS Treatment of locoregionally advanced head and neck cancer with concomitant highdose radiotherapy plus cetuximab improves locoregional control and reduces mortality without increasing the common toxic effects associated with radiotherapy to the head and neck. (ClinicalTrials.gov number, NCT00004227.)

A unified mixed-model method for association mapping that accounts for multiple levels of relatedness

Abstract: As population structure can result in spurious associations, it has constrained the use of association studies in human and plant genetics. Association mapping, however, holds great promise if true signals of functional association can be separated from the vast number of false signals generated by population structure. We have developed a unified mixed-model approach to account for multiple levels of relatedness simultaneously as detected by random genetic markers. We applied this new approach to two samples: a family-based sample of 14 human families, for quantitative gene expression dissection, and a sample of 277 diverse maize inbred lines with complex familial relationships and population structure, for quantitative trait dissection. Our method demonstrates improved control of both type I and type II error rates over other methods. As this new method crosses the boundary between family-based and structured association samples, it provides a powerful complement to currently available methods for association mapping.

Common Method Variance in IS Research: A Comparison of Alternative Approaches and a Reanalysis of Past Research

Abstract: Despite recurring concerns about common method variance (CMV) in survey research, the information systems (IS) community remains largely uncertain of the extent of such potential biases. To address this uncertainty, this paper attempts to systematically examine the impact of CMV on the inferences drawn from survey research in the IS area. First, we describe the available approaches for assessing CMV and conduct an empirical study to compare them. From an actual survey involving 227 respondents, we find that although CMV is present in the research areas examined, such biases are not substantial. The results also suggest that few differences exist between the relatively new marker-variable technique and other well-established conventional tools in terms of their ability to detect CMV. Accordingly, the marker-variable technique was employed to infer the effect of CMV on correlations from previously published studies. Our findings, based on the reanalysis of 216 correlations, suggest that the inflated correlation caused by CMV may be expected to be on the order of 0.10 or less, and most of the originally significant correlations remain significant even after controlling for CMV. Finally, by extending the marker-variable technique, we examined the effect of CMV on structural relationships in past literature. Our reanalysis reveals that contrary to the concerns of some skeptics, CMV-adjusted structural relationships not only remain largely significant but also are not statistically differentiable from uncorrected estimates. In summary, this comprehensive and systematic analysis offers initial evidence that (1) the marker-variable technique can serve as a convenient, yet effective, tool for accounting for CMV, and (2) common method biases in the IS domain are not as serious as those found in other disciplines.

From the Achievement Gap to the Education Debt: Understanding Achievement in U.S. Schools

Abstract: The achievement gap is one of the most talked-about issues in U.S. education. The term refers to the disparities in standardized test scores between Black and White, Latina/o and White, and recent immigrant and White students. This article argues that a focus on the gap is misplaced. Instead, we need to look at the “education debt” that has accumulated over time. This debt comprises historical, economic, sociopolitical, and moral components. The author draws an analogy with the concept of national debt—which she contrasts with that of a national budget deficit—to argue the significance of the education debt.

Calcium plus vitamin D supplementation and the risk of fractures.

Abstract: Background The efficacy of calcium with vitamin D supplementation for preventing hip and other fractures in healthy postmenopausal women remains equivocal. Methods We recruited 36,282 postmenopausal women, 50 to 79 years of age, who were already enrolled in a Women's Health Initiative (WHI) clinical trial. We randomly assigned participants to receive 1000 mg of elemental calcium as calcium carbonate with 400 IU of vitamin D3 daily or placebo. Fractures were ascertained for an average follow-up period of 7.0 years. Bone density was measured at three WHI centers. Results Hip bone density was 1.06 percent higher in the calcium plus vitamin D group than in the placebo group (P<0.01). Intention-to-treat analysis indicated that participants receiving calcium plus vitamin D supplementation had a hazard ratio of 0.88 for hip fracture (95 percent confidence interval, 0.72 to 1.08), 0.90 for clinical spine fracture (0.74 to 1.10), and 0.96 for total fractures (0.91 to 1.02). The risk of renal calculi increased with...

Financial Constraints Risk

Abstract: We construct an index of firms’ external finance constraints via generalized method of moments (GMM) estimation of an investment Euler equation. Unlike the commonly used KZ index, ours is consistent with firm characteristics associated with external finance constraints. Constrained firms’ returns move together, suggesting the existence of a financial constraints factor. This factor earns a positive but insignificant average return. Much of the variation in this factor cannot be explained by the Fama–French and momentum factors. Cross-sectional regressions of returns on our index and other firm characteristics show that constrained firms earn higher returns and that the financial-constraints effect dominates the size effect. We explore the impact of firms’ external finance constraints on their stock returns. Motivation for this inquiry starts with a large body of microeconometric studies that have provided some evidence of an impact of external finance constraints on investment. For example, Whited (1992), Bond and Meghir (1994), and Love (2003) show that augmentations of an investment Euler equation that account for financial constraints improve its fit. The question remains whether these effects are priced in asset markets. In other words, do financial constraints affect asset returns; and if so, is this risk diversifiable? To tackle this question, we construct an index of financial constraints based on a standard intertemporal investment model augmented to account for financial frictions. The model predicts that external finance constraints affect the intertemporal substitution of investment today for investment tomorrow via the shadow value of scarce external funds. This shadow value in turn depends on observable variables. Generalized method of moments (GMM) estimation of the model provides fitted values of the shadow value, which we then use as our index. The most important advantage of this approach is its avoidance of serious sample selection, simultaneity, and measurement-error problems via structural

Collagen reorganization at the tumor-stromal interface facilitates local invasion

Abstract: Stromal-epithelial interactions are of particular significance in breast tissue as misregulation of these interactions can promote tumorigenesis and invasion. Moreover, collagen-dense breast tissue increases the risk of breast carcinoma, although the relationship between collagen density and tumorigenesis is not well understood. As little is known about epithelial-stromal interactions in vivo, it is necessary to visualize the stroma surrounding normal epithelium and mammary tumors in intact tissues to better understand how matrix organization, density, and composition affect tumor formation and progression. Epithelial-stromal interactions in normal mammary glands, mammary tumors, and tumor explants in three-dimensional culture were studied with histology, electron microscopy, and nonlinear optical imaging methodologies. Imaging of the tumor-stromal interface in live tumor tissue ex vivo was performed with multiphoton laser-scanning microscopy (MPLSM) to generate multiphoton excitation (MPE) of endogenous fluorophores and second harmonic generation (SHG) to image stromal collagen. We used both laser-scanning multiphoton and second harmonic generation microscopy to determine the organization of specific collagen structures around ducts and tumors in intact, unfixed and unsectioned mammary glands. Local alterations in collagen density were clearly seen, allowing us to obtain three-dimensional information regarding the organization of the mammary stroma, such as radiating collagen fibers that could not have been obtained using classical histological techniques. Moreover, we observed and defined three tumor-associated collagen signatures (TACS) that provide novel markers to locate and characterize tumors. In particular, local cell invasion was found predominantly to be oriented along certain aligned collagen fibers, suggesting that radial alignment of collagen fibers relative to tumors facilitates invasion. Consistent with this observation, primary tumor explants cultured in a randomly organized collagen matrix realigned the collagen fibers, allowing individual tumor cells to migrate out along radially aligned fibers. The presentation of these tumor-associated collagen signatures allowed us to identify pre-palpable tumors and see cells at the tumor-stromal boundary invading into the stroma along radially aligned collagen fibers. As such, TACS should provide indications that a tumor is, or could become, invasive, and may serve as part of a strategy to help identify and characterize breast tumors in animal and human tissues.

The economics of poverty traps and persistent poverty: An asset-based approach

Abstract: Longitudinal data on household living standards open the way to a deeper analysis of the nature and extent of poverty. While a number of studies have exploited this type of data to distinguish transitory from more chronic forms of income or expenditure poverty, this paper develops an asset-based approach to poverty analysis that makes it possible to distinguish deep-rooted, persistent structural poverty from poverty that passes naturally with time due to systemic growth processes. Drawing on the economic theory of poverty traps and bifurcated accumulation strategies, this paper briefly discusses some feasible estimation strategies for empirically identifying poverty traps and long-term, persistent structural poverty, as well as relevant extensions of the popular Foster-Greer-Thorbecke class of poverty measures. The paper closes with reflections on how asset-based poverty can be used to underwrite the design of persistent poverty reduction strategies.

Phase Modifiers Promote Efficient Production of Hydroxymethylfurfural from Fructose

Abstract: Furan derivatives obtained from renewable biomass resources have the potential to serve as substitutes for the petroleum-based building blocks that are currently used in the production of plastics and fine chemicals. We developed a process for the selective dehydration of fructose to 5-hydroxymethylfurfural (HMF) that operates at high fructose concentrations (10 to 50 weight %), achieves high yields (80% HMF selectivity at 90% fructose conversion), and delivers HMF in a separation-friendly solvent. In a two-phase reactor system, fructose is dehydrated in the aqueous phase with the use of an acid catalyst (hydrochloric acid or an acidic ion-exchange resin) with dimethylsulfoxide and/or poly(1-vinyl-2-pyrrolidinone) added to suppress undesired side reactions. The HMF product is continuously extracted into an organic phase (methylisobutylketone) modified with 2-butanol to enhance partitioning from the reactive aqueous solution.

Efficacy of Varenicline, an 42 Nicotinic Acetylcholine Receptor Partial Agonist, vs Placebo or Sustained-Release Bupropion for Smoking Cessation A Randomized Controlled Trial

Abstract: ContextVarenicline, a partial agonist at the α4β2 nicotinic acetylcholine receptor, has the potential to aid smoking cessation by relieving nicotine withdrawal symptoms and reducing the rewarding properties of nicotine.ObjectiveTo determine the efficacy and safety of varenicline for smoking cessation compared with placebo or sustained-release bupropion (bupropion SR).Design, Setting, and ParticipantsA randomized, double-blind, placebo-controlled trial conducted between June 2003 and March 2005 at 14 research centers with a 12-week treatment period and follow-up of smoking status to week 52. Of 1413 adult smokers who volunteered for the study, 1027 were enrolled; 65% of randomized participants completed the study.InterventionVarenicline titrated to 1 mg twice daily (n = 344) or bupropion SR titrated to 150 mg twice daily (n = 342) or placebo (n = 341) for 12 weeks, plus weekly brief smoking cessation counseling.Main Outcome MeasuresContinuous abstinence from smoking during the last 4 weeks of treatment (weeks 9-12; primary end point) and through the follow-up period (weeks 9-24 and 9-52).ResultsDuring the last 4 weeks of treatment (weeks 9-12), 43.9% of participants in the varenicline group were continuously abstinent from smoking compared with 17.6% in the placebo group (odds ratio [OR], 3.85; 95% confidence interval [CI], 2.69-5.50; P<.001) and 29.8% in the bupropion SR group (OR, 1.90; 95% CI, 1.38-2.62; P<.001). For weeks 9 through 24, 29.7% of participants in the varenicline group were continuously abstinent compared with 13.2% in the placebo group (OR, 2.83; 95% CI, 1.91-4.19; P<.001) and 20.2% in the bupropion group (OR, 1.69; 95% CI, 1.19-2.42; P = .003). For weeks 9 through 52, 23% of participants in the varenicline group were continuously abstinent compared with 10.3% in the placebo group (OR, 2.66; 95% CI, 1.72-4.11; P<.001) and 14.6% in the bupropion SR group (OR, 1.77; 95% CI, 1.19-2.63; P = .004). Treatment was discontinued due to adverse events by 10.5% of participants in the varenicline group, 12.6% in the bupropion SR group, and 7.3% in the placebo group. The most common adverse event with varenicline was nausea, which occurred in 101 participants (29.4%).ConclusionsVarenicline is an efficacious, safe, and well-tolerated smoking cessation pharmacotherapy. Varenicline's short-term and long-term efficacy exceeded that of both placebo and bupropion SR.Trial Registrationclinicaltrials.gov Identifier: NCT00143364

Asymmetric and symmetric stem-cell divisions in development and cancer

Abstract: Much has been made of the idea that asymmetric cell division is a defining characteristic of stem cells that enables them to simultaneously perpetuate themselves (self-renew) and generate differentiated progeny. Yet many stem cells can divide symmetrically, particularly when they are expanding in number during development or after injury. Thus, asymmetric division is not necessary for stem-cell identity but rather is a tool that stem cells can use to maintain appropriate numbers of progeny. The facultative use of symmetric or asymmetric divisions by stem cells may be a key adaptation that is crucial for adult regenerative capacity.

Reactive oxygen species: role in the development of cancer and various chronic conditions

Abstract: Oxygen derived species such as superoxide radical, hydrogen peroxide, singlet oxygen and hydroxyl radical are well known to be cytotoxic and have been implicated in the etiology of a wide array of human diseases, including cancer. Various carcinogens may also partly exert their effect by generating reactive oxygen species (ROS) during their metabolism. Oxidative damage to cellular DNA can lead to mutations and may, therefore, play an important role in the initiation and progression of multistage carcinogenesis. The changes in DNA such as base modification, rearrangement of DNA sequence, miscoding of DNA lesion, gene duplication and the activation of oncogenes may be involved in the initiation of various cancers. Elevated levels of ROS and down regulation of ROS scavengers and antioxidant enzymes are associated with various human diseases including various cancers. ROS are also implicated in diabtes and neurodegenerative diseases. ROS influences central cellular processes such as proliferation a, apoptosis, senescence which are implicated in the development of cancer. Understanding the role of ROS as key mediators in signaling cascades may provide various opportunities for pharmacological intervention.

Comparative genomics of the lactic acid bacteria

Abstract: Lactic acid-producing bacteria are associated with various plant and animal niches and play a key role in the production of fermented foods and beverages. We report nine genome sequences representing the phylogenetic and functional diversity of these bacteria. The small genomes of lactic acid bacteria encode a broad repertoire of transporters for efficient carbon and nitrogen acquisition from the nutritionally rich environments they inhabit and reflect a limited range of biosynthetic capabilities that indicate both prototrophic and auxotrophic strains. Phylogenetic analyses, comparison of gene content across the group, and reconstruction of ancestral gene sets indicate a combination of extensive gene loss and key gene acquisitions via horizontal gene transfer during the coevolution of lactic acid bacteria with their habitats.

Avian flu: influenza virus receptors in the human airway.

Abstract: Avian and human flu viruses seem to target different regions of a patient's respiratory tract.

The Risk of Bloodstream Infection in Adults With Different Intravascular Devices: A Systematic Review of 200 Published Prospective Studies

Abstract: OBJECTIVE: To better understand the absolute and relative risks of bloodstream infection (BSI) associated with the various types of intravascular devices (IVDs), we analyzed 200 published studies of adults in which every device in the study population was prospectively evaluated for evidence of associated infection and microbiologically based criteria were used to define IVD-related BSI. METHODS: English-language reports of prospective studies of adults published between January 1, 1966, and July 1, 2005, were identified by MEDLINE search using the following general search strategy: bacteremia [Medical Subject Heading, MeSH] OR septicemia [MeSH] OR bloodstream infection AND the specific type of intravascular device (eg, central venous port). Mean rates of IVDrelated BSI were calculated from pooled data for each type of device and expressed as BSIs per 100 IVDs (%) and per 1000 IVD days. RESULTS: Point incidence rates of IVD-related BSI were lowest with peripheral intravenous catheters (0.1%, 0.5 per 1000 IVDdays) and midline catheters (0.4%, 0.2 per 1000 catheter-days). Far higher rates were seen with short-term noncuffed and nonmedicated central venous catheters (CVCs) (4.4%, 2.7 per 1000 catheter-days). Arterial catheters used for hemodynamic monitoring (0.8%, 1.7 per 1000 catheter-days) and peripherally inserted central catheters used in hospitalized patients (2.4%, 2.1 per 1000 catheter-days) posed risks approaching those seen with short-term conventional CVCs used in the intensive care unit. Surgically implanted long-term central venous devices—cuffed and tunneled catheters (22.5%, 1.6 per 1000 IVD-days) and central venous ports (3.6%, 0.1 per 1000 IVD-days)—appear to have high rates of infection when risk is expressed as BSIs per 100 IVDs but actually pose much lower risk when rates are expressed per 1000 IVD-days. The use of cuffed and tunneled dual lumen CVCs rather than noncuffed, nontunneled catheters for temporary hemodialysis and novel preventive technologies, such as CVCs with anti-infective surfaces, was associated with considerably lower rates of catheter-related BSI. CONCLUSIONS: Expressing risk of IVD-related BSI per 1000 IVDdays rather than BSIs per 100 IVDs allows for more meaningful estimates of risk. These data, based on prospective studies in which every IVD in the study cohort was analyzed for evidence of infection by microbiologically based criteria, show that all types of IVDs pose a risk of IVD-related BSI and can be used for benchmarking rates of infection caused by the various types of IVDs in use at the present time. Since almost all the national effort and progress to date to reduce the risk of IVD-related infection have focused on short-term noncuffed CVCs used in intensive care units, infection control programs must now strive to consistently apply essential control measures and preventive technologies with all types of IVDs.

Work system design for patient safety: the SEIPS model

Abstract: Models and methods of work system design need to be developed and implemented to advance research in and design for patient safety. In this paper we describe how the Systems Engineering Initiative for Patient Safety (SEIPS) model of work system and patient safety, which provides a framework for understanding the structures, processes and outcomes in health care and their relationships, can be used toward these ends. An application of the SEIPS model in one particular care setting (outpatient surgery) is presented and other practical and research applications of the model are described.

A handful of heuristics and some propositions for understanding resilience in social-ecological systems

Abstract: This paper is a work-in-progress account of ideas and propositions about resilience in socialecological systems. It articulates our understanding of how these complex systems change and what determines their ability to absorb disturbances in either their ecological or their social domains. We call them “propositions” because, although they are useful in helping us understand and compare different social-ecological systems, they are not sufficiently well defined to be considered formal hypotheses. These propositions were developed in two workshops, in 2003 and 2004, in which participants compared the dynamics of 15 case studies in a wide range of regions around the world. The propositions raise many questions, and we present a list of some that could help define the next phase of resilience-related research.

Gene–environment interactions in psychiatry: joining forces with neuroscience

Abstract: Gene-environment interaction research in psychiatry is new, and is a natural ally of neuroscience. Mental disorders have known environmental causes, but there is heterogeneity in the response to each causal factor, which gene-environment findings attribute to genetic differences at the DNA sequence level. Such findings come from epidemiology, an ideal branch of science for showing that a gene-environment interactions exist in nature and affect a significant fraction of disease cases. The complementary discipline of epidemiology, experimental neuroscience, fuels gene-environment hypotheses and investigates underlying neural mechanisms. This article discusses opportunities and challenges in the collaboration between psychiatry, epidemiology and neuroscience in studying gene-environment interactions.

Derivation of human embryonic stem cells in defined conditions.

Abstract: We have previously reported that high concentrations of basic fibroblast growth factor (bFGF) support feeder-independent growth of human embryonic stem (ES) cells, but those conditions included poorly defined serum and matrix components. Here we report feeder-independent human ES cell culture that includes protein components solely derived from recombinant sources or purified from human material. We describe the derivation of two new human ES cell lines in these defined culture conditions.

Electrical control of antiferromagnetic domains in multiferroic BiFeO3 films at room temperature.

Abstract: Multiferroic materials, which offer the possibility of manipulating the magnetic state by an electric field or vice versa, are of great current interest. In this work, we demonstrate the first observation of electrical control of antiferromagnetic domain structure in a single-phase multiferroic material at room temperature. High-resolution images of both antiferromagnetic and ferroelectric domain structures of (001)-oriented multiferroic BiFeO3 films revealed a clear domain correlation, indicating a strong coupling between the two types of order. The ferroelectric structure was measured using piezo force microscopy, whereas X-ray photoemission electron microscopy as well as its temperature dependence was used to detect the antiferromagnetic configuration. Antiferromagnetic domain switching induced by ferroelectric polarization switching was observed, in agreement with theoretical predictions.

Mondrian Multidimensional K-Anonymity

Abstract: K-Anonymity has been proposed as a mechanism for protecting privacy in microdata publishing, and numerous recoding "models" have been considered for achieving 𝑘anonymity This paper proposes a new multidimensional model, which provides an additional degree of flexibility not seen in previous (single-dimensional) approaches Often this flexibility leads to higher-quality anonymizations, as measured both by general-purpose metrics and more specific notions of query answerability Optimal multidimensional anonymization is NP-hard (like previous optimal 𝑘-anonymity problems) However, we introduce a simple greedy approximation algorithm, and experimental results show that this greedy algorithm frequently leads to more desirable anonymizations than exhaustive optimal algorithms for two single-dimensional models