Multiple Additive Regression Tree Research Articles

Mediators of racial/ethnic inequities in clinical trial participation among patients with cancer, 2011-2023.

While minoritized populations are less likely to participate in cancer trials, it is unknown whether social determinants of health (SDOH) explain these inequities. Here we identify SDOH factors that contribute to racial/ethnic inequities in clinical trial participation among patients with 22 common cancers. This retrospective cohort study used electronic health record data (2011-2023) linked to neighborhood (Census tract) data from multiple sources. Patients were followed from diagnosis to clinical study drug receipt (proxy for trial participation), death, or last recorded activity. Associations were assessed using Cox proportional hazards models adjusted for clinical factors (diagnosis year, age, sex, performance status, stage, cancer type). To elucidate which area-level SDOH underlie racial/ethnic inequities, mediation analysis was performed using nonlinear multiple additive regression tree models. This study included 250105 patients (64.7% non-Latinx White, 8.9% non-Latinx Black, 5.2% Latinx). Black and Latinx patients were more likely to live in economically/socially marginalized areas (eg, disproportionately minoritized [measure of segregation], limited English proficiency [LEP], low vehicle ownership) than White patients. Black (3.7%; HR = 0.55 [CI = 0.52-0.60]) and Latinx patients (4.4%; HR = 0.63 [CI = 0.58-0.69]) were less likely to participate in trials than White patients (6.3%). Fewer patients in economically/socially marginalized neighborhoods participated in trials. Mediators explained 62.2% (CI = 49.5%-74.8%) of participation inequities between Black and White patients; area-level SDOH-including segregation (29.9% [CI = 21.2%-38.6%]) and vehicle ownership (11.6% [CI = 7.0%-16.1%])-were the most important mediators. Similarly, Latinx-White participation inequities were mediated (65.1%, [CI = 49.8%-80.3%]) by area-level SDOH such as segregation (39.8% [CI = 28.3%-51.3%]), LEP (11.6%, [CI = 2.8%-20.4%]), and vehicle ownership (9.6% [CI = 5.8%-13.5%]). To improve racial/ethnic diversity in cancer trials, efforts to address barriers related to adverse neighborhood SDOH factors are necessary.

Multi-Level Factors Underlying Racial/Ethnic Inequities in Clinical Trial Participation Among Patients with Hematologic Cancers: Lessons for the Development of Diversity Plans

Background: The US Food & Drug Administration announced that drug sponsors will be required to implement plans to improve racial/ethnic diversity in clinical trials starting in 2024. While historically marginalized racial/ethnic populations are less likely to participate in cancer trials, there is limited information regarding inequities in access to hematologic cancer trials. It is also unknown whether social determinants of health (SDOH) explain racial/ethnic inequities in trial participation. Here we examined the extent to which commonly used trial inclusion criteria and SDOH factors explain racial/ethnic inequities in clinical trial participation among patients with six hematologic cancers. Methods: This retrospective study used the nationwide Flatiron Health electronic health record-derived de-identified database (2011-2022) of adult patients with multiple myeloma, non-Hodgkin's lymphoma (diffuse large B-cell, follicular, or mantle cell lymphoma) or leukemia (chronic lymphocytic or acute myeloid leukemia). Area-level SDOH ( e.g. predominant race/ethnicity - a measure of segregation) characterizing where patients lived were assessed using the American Community Survey census tract data. Patients were followed from initial diagnosis to clinical study drug receipt (a proxy for trial participation), death, or last recorded activity. Associations of race/ethnicity and SDOH factors with trial participation were assessed using cause-specific hazards models adjusted for clinical confounders (diagnosis year, age, sex, and cancer type). To determine which commonly used trial inclusion criteria ( i.e. abnormal creatinine serum [≥ 2] and performance status) and SDOH factors underlie racial/ethnic inequities, mediation analysis was performed using nonlinear multiple additive regression tree models. Results: This study included 51,382 patients (median age: 67 years, 42.2% women; 57.7% non-Latinx [NL] White, 6.9% NL Black, 6.1% Latinx, 29.4% other/unknown). Black and Latinx patients were more likely to live in marginalized areas ( e.g. segregated neighborhoods color and higher percentages of limited English proficiency and low vehicle ownership) than White patients. Overall, 5.6% of patients participated in trials. Patients who received care from community practices or lived in marginalized neighborhoods were less likely to participate in trials. Black (4.2%; hazard ratio [HR] 0.62, 95% confidence interval [CI]: 0.52- 0.73) and Latinx patients (3.7%; HR 0.53, CI: 0.44-0.65) were less likely to participate in trials than White patients (6.4%). Black-White trial inequities were mediated (77%, CI: 42%-117%) by practice type ( e.g., community vs. academic center, 19%, CI: 9%-28%) and area-level SDOH such as segregation (49%, CI: 21%-77%) and vehicle ownership (12%, CI: 1%-24%). Mediators explained 55% (CI: 23%-87%) of the Latinx-White trial participation inequity; area-level SDOH-including segregation (30%, CI: 3%-58%), limited English proficiency (21%, CI: -3% to 45%), and vehicle ownership (17%, CI: 0%-45%)-were the most important mediators. Clinical factors did not substantially mediate racial/ethnic inequities (< 3% proportion mediated). Conclusion: Black and Latinx patients diagnosed with hematologic cancers are 40% to 50% less likely to participate in clinical trials than their White counterparts. SDOH factors explained most of these inequities. These findings suggest that efforts to increase racial/ethnic diversity in cancer trials must address the persistent effects of segregation, along with everyday obstacles to clinical care, such as transportation and language barriers. Without concerted efforts to increase trial participation among historically marginalized racial/ethnic groups, trials may perpetuate inequities in investigational drug access and adversely impact the generalizability of pivotal cancer treatment trials.

Analysis of Race and Ethnicity, Socioeconomic Factors, and Tooth Decay Among US Children

ImportanceWhile large oral health disparities remain by race and ethnicity among children, the associations of race, ethnicity, and mediating factors with oral health outcomes are poorly characterized. Identifying the pathways that explain these disparities would be critical to inform policies to effectively reduce them.ObjectiveTo measure racial and ethnic disparities in the risk of developing tooth decay and quantify relative contributions of factors mediating the observed disparities among US children.Design, Setting, and ParticipantsThis retrospective cohort study used electronic health records of US children from 2014 to 2020 to measure racial and ethnic disparities in the risk of tooth decay. Elastic net regularization was used to select variables to be included in the model among medical conditions, dental procedure types, and individual- and community-level socioeconomic factors. Data were analyzed from January 9 to April 28, 2023.ExposuresRace and ethnicity of children.Main Outcomes and MeasuresThe main outcome was diagnosis of tooth decay in either deciduous or permanent teeth, defined as at least 1 decayed, filled, or missing tooth due to caries. An Anderson-Gill model, a time-to-event model for recurrent tooth decay events with time-varying covariates, stratified by age groups (0-5, 6-10, and 11-18 years) was estimated. A nonlinear multiple additive regression tree–based mediation analysis quantified the relative contributions of factors underlying the observed racial and ethnic disparities.ResultsAmong 61 083 children and adolescents aged 0 to 18 years at baseline (mean [SD] age, 9.9 [4.6] years; 30 773 [50.4%] female), 2654 Black individuals (4.3%), 11 213 Hispanic individuals (18.4%), 42 815 White individuals (70.1%), and 4401 individuals who identified as another race (eg, American Indian, Asian, and Hawaiian and Pacific Islander) (7.2%) were identified. Larger racial and ethnic disparities were observed among children aged 0 to 5 years compared with other age groups (Hispanic children: adjusted hazard ratio [aHR], 1.47; 95% CI, 1.40-1.54; Black children: aHR, 1.30; 95% CI, 1.19-1.42; other race children: aHR, 1.39; 95% CI, 1.29-1.49), compared with White children. For children aged 6 to 10 years, higher risk of tooth decay was observed for Black children (aHR, 1.09; 95% CI, 1.01-1.19) and Hispanic children (aHR, 1.12; 95% CI, 1.07-1.18) compared with White children. For adolescents aged 11 to 18 years, a higher risk of tooth decay was observed only in Black adolescents (aHR, 1.17; 95% CI, 1.06-1.30). A mediation analysis revealed that the association of race and ethnicity with time to first tooth decay became negligible, except for Hispanic and children of other race aged 0 to 5 years, suggesting that mediators explained most of the observed disparities. Insurance type explained the largest proportion of the disparity, ranging from 23.4% (95% CI, 19.8%-30.2%) to 78.9% (95% CI, 59.0%-114.1%), followed by dental procedures (receipt of topical fluoride and restorative procedures) and community-level factors (education attainment and Area Deprivation Index).ConclusionsIn this retrospective cohort study, large proportions of disparities in time to first tooth decay associated with race and ethnicity were explained by insurance type and dental procedure types among children and adolescents. These findings can be applied to develop targeted strategies to reduce oral health disparities.

Is the technical quality of root fillings associated with socioeconomic status?

Abstract Background Socioeconomically better-off tend to receive better care, and to benefit more from care they receive, than the less well-off. The technical quality of root fillings can be assessed radiographically. Our aim was to investigate differences in the technical quality of root fillings according to socioeconomic status not studied before. Methods In the Finnish Health 2000 Survey, a sample of 8 028 over 29-year-old adults was selected by a stratified two-staged cluster sampling. Of these, 6 115 (76%) underwent panoramic radiography. A total of 7 986 endodontically treated teeth were detected among 3 905 participants. The criterion for a technically inadequate root filling was a distance from the root filling to the root apex of over 3 mm or material over the apex. Education, employment status and household income were used as measures of socioeconomic status. Differences in the quality of root fillings according to socioeconomic status were investigated using the Multiple Additive Regression Tree mediation analysis, taking into account age, gender and tooth type of the root filled tooth. Results Almost half (47%) of the root fillings were inadequate. Root filling quality was associated particularly with age of the participant and type of tooth. Mediation analysis showed that among the over 64-year-olds, higher education or income were associated with a slightly higher probability of inadequate root filling quality (ORs 1.15-1.20, p < 0.05). Among the under 65-year-olds, age of the participant and type of tooth explained the observed variations in root filling quality according to socieconomic status, and socioeconomic status had no influence (ORs 0.99-1.02, p > 0.05). Conclusions Those in a lower socioeconomic status had not received poorer endodontic treatment quality. It seems that those in a lower socioeconomic status are not prone to receive poorer quality care in endodontic setting, that they tend to receive in longer and more socially complex treatments. Key messages Variations in the technical quality of root fillings by socioeconomic status has not been studied. Lower socioeconomic status was not associated with poorer technical quality of root fillings.

Meta-XGBoost for Hyperspectral Image Classification Using Extended MSER-Guided Morphological Profiles

To investigate the performance of extreme gradient boosting (XGBoost) in remote sensing image classification tasks, XGBoost was first introduced and comparatively investigated for the spectral-spatial classification of hyperspectral imagery using the extended maximally stable extreme-region-guided morphological profiles (EMSER_MPs) proposed in this study. To overcome the potential issues of XGBoost, meta-XGBoost was proposed as an ensemble XGBoost method with classification and regression tree (CART), dropout-introduced multiple additive regression tree (DART), elastic net regression and parallel coordinate descent-based linear regression (linear) and random forest (RaF) boosters. Moreover, to evaluate the performance of the introduced XGBoost approach with different boosters, meta-XGBoost and EMSER_MPs, well-known and widely accepted classifiers, including support vector machine (SVM), bagging, adaptive boosting (AdaBoost), multi class AdaBoost (MultiBoost), extremely randomized decision trees (ExtraTrees), RaF, classification via random forest regression (CVRFR) and ensemble of nested dichotomies with extremely randomized decision tree (END-ERDT) methods, were considered in terms of the classification accuracy and computational efficiency. The experimental results based on two benchmark hyperspectral data sets confirm the superior performance of EMSER_MPs and EMSER_MPs with mean pixel values within region (EMSER_MPsM) compared to that for morphological profiles (MPs), morphological profile with partial reconstruction (MPPR), extended MPs (EMPs), extended MPPR (EMPPR), maximally stable extreme-region-guided morphological profiles (MSER_MPs) and MSER_MPs with mean pixel values within region (MSER_MPsM) features. The proposed meta-XGBoost algorithm is capable of obtaining better results than XGBoost with the CART, DART, linear and RaF boosters, and it could be an alternative to the other considered classifiers in terms of the classification of hyperspectral images using advanced spectral-spatial features, especially from generalized classification accuracy and model training efficiency perspectives.

Twenty-five-year follow-up of educational differences in toothache prevalence.

Nationally representative studies on socioeconomic differences in toothache prevalence are very limited, especially with a long follow-up period. Our aim was to examine the education-related differences in toothache prevalence among 25- to 64-year-olds in Finland during 1990-2014. We used data from the Health Behaviour and Health Among the Finnish Adult Population surveys. Data were gathered by random samples of 15- to 64-year-old Finns annually between 1990 and 2014. Response rates decreased during this period from 76% (n=3812) to 53% (n=2630). Our final pooled total sample size was 63372 after exclusion of edentate participants and missing data. Relative educational level was used to measure socioeconomic position. In addition to basic tabulations, educational differences in toothache prevalence during the past month were investigated using the multiple additive regression tree for mediation analysis. The proportion of respondents who reported having had toothache during the past month increased from 7% to 12% from 1990 to 2014. Educational differences in toothache prevalence were small for the whole study period, especially in women. Men with high relative education had a lower toothache prevalence than people with low relative education during 1990-1994 and 2010-2014. This stemmed from the direct effect of relative education and the indirect effects of toothbrushing frequency, number of missing teeth and perceived health. In both sexes, those with middle-level relative education had higher toothache prevalence than people with low relative education 1995-2009. During 1990-2014, toothache prevalence increased among 25- to 64-year-olds in Finland. Since the mid-1990s, differences in toothache prevalence by relative education remained very small in both sexes. More actions are needed to understand reasons behind the clearly increased toothache prevalence in Finland during the last decades.

Distributed Deep Forest and its Application to Automatic Detection of Cash-Out Fraud

Internet companies are facing the need for handling large-scale machine learning applications on a daily basis and distributed implementation of machine learning algorithms which can handle extra-large-scale tasks with great performance is widely needed. Deep forest is a recently proposed deep learning framework which uses tree ensembles as its building blocks and it has achieved highly competitive results on various domains of tasks. However, it has not been tested on extremely large-scale tasks. In this work, based on our parameter server system, we developed the distributed version of deep forest. To meet the need for real-world tasks, many improvements are introduced to the original deep forest model, including MART (Multiple Additive Regression Tree) as base learners for efficiency and effectiveness consideration, the cost-based method for handling prevalent class-imbalanced data, MART based feature selection for high dimension data, and different evaluation metrics for automatically determining the cascade level. We tested the deep forest model on an extra-large-scale task, i.e., automatic detection of cash-out fraud, with more than 100 million training samples. Experimental results showed that the deep forest model has the best performance according to the evaluation metrics from different perspectives even with very little effort for parameter tuning. This model can block fraud transactions in a large amount of money each day. Even compared with the best-deployed model, the deep forest model can additionally bring a significant decrease in economic loss each day.

Analysis of Tourist Flow Forecasting Model Based on Multiple Additive Regression Tree

Accurate prediction of tourist flow is a key issue in tourism economic analysis and development planning. This paper proposes a tourist flow prediction model based on Multiple Additive Regression Tree (MART) by using machine learning ideas. The model uses factors such as temperature, sunshine duration, air quality and so on to construct eigenvector, and constructing multiple base learners through the boosting framework to predict tourist flow accurately. Take Guilin city from 2015 to 2018 Tourist as an example for analysis, the prediction accuracy of the model is evaluated by means of the average error, square equalization error and other indicators. The experimental results show that the proposed method has high accuracy in the prediction of tourist flow.

Into the Bowels of Depression: Unravelling Medical Symptoms Associated with Depression by Applying Machine-Learning Techniques to a Community Based Population Sample.

BackgroundDepression is commonly comorbid with many other somatic diseases and symptoms. Identification of individuals in clusters with comorbid symptoms may reveal new pathophysiological mechanisms and treatment targets. The aim of this research was to combine machine-learning (ML) algorithms with traditional regression techniques by utilising self-reported medical symptoms to identify and describe clusters of individuals with increased rates of depression from a large cross-sectional community based population epidemiological study.MethodsA multi-staged methodology utilising ML and traditional statistical techniques was performed using the community based population National Health and Nutrition Examination Study (2009–2010) (N = 3,922). A Self-organised Mapping (SOM) ML algorithm, combined with hierarchical clustering, was performed to create participant clusters based on 68 medical symptoms. Binary logistic regression, controlling for sociodemographic confounders, was used to then identify the key clusters of participants with higher levels of depression (PHQ-9≥10, n = 377). Finally, a Multiple Additive Regression Tree boosted ML algorithm was run to identify the important medical symptoms for each key cluster within 17 broad categories: heart, liver, thyroid, respiratory, diabetes, arthritis, fractures and osteoporosis, skeletal pain, blood pressure, blood transfusion, cholesterol, vision, hearing, psoriasis, weight, bowels and urinary.ResultsFive clusters of participants, based on medical symptoms, were identified to have significantly increased rates of depression compared to the cluster with the lowest rate: odds ratios ranged from 2.24 (95% CI 1.56, 3.24) to 6.33 (95% CI 1.67, 24.02). The ML boosted regression algorithm identified three key medical condition categories as being significantly more common in these clusters: bowel, pain and urinary symptoms. Bowel-related symptoms was found to dominate the relative importance of symptoms within the five key clusters.ConclusionThis methodology shows promise for the identification of conditions in general populations and supports the current focus on the potential importance of bowel symptoms and the gut in mental health research.

Predicting defect behavior in B2 intermetallics by merging ab initio modeling and machine learning

AbstractWe present a combination of machine learning and high throughput calculations to predict the points defects behavior in binary intermetallic (A–B) compounds, using as an example systems with the cubic B2 crystal structure (with equiatomic AB stoichiometry). To the best of our knowledge, this work is the first application of machine learning-models for point defect properties. High throughput first principles density functional calculations have been employed to compute intrinsic point defect energies in 100 B2 intermetallic compounds. The systems are classified into two groups: (i) those for which the intrinsic defects are antisites for both A and B rich compositions, and (ii) those for which vacancies are the dominant defect for either or both composition ranges. The data was analyzed by machine learning-techniques using decision tree, and full and reduced multiple additive regression tree (MART) models. Among these three schemes, a reduced MART (r-MART) model using six descriptors (formation energy, minimum and difference of electron densities at the Wigner–Seitz cell boundary, atomic radius difference, maximal atomic number and maximal electronegativity) presents the highest fit (98 %) and predictive (75 %) accuracy. This model is used to predict the defect behavior of other B2 compounds, and it is found that 45 % of the compounds considered feature vacancies as dominant defects for either A or B rich compositions (or both). The ability to predict dominant defect types is important for the modeling of thermodynamic and kinetic properties of intermetallic compounds, and the present results illustrate how this information can be derived using modern tools combining high throughput calculations and data analytics.

Predicting alien herb invasion with machine learning models: biogeographical and life-history traits both matter

Identifying the variables associated with invasiveness is a core task for developing risk assessment models to predict invasion potential. However, quantitative models with both biogeographical and life-history variables for invasion risk assessment in China are limited. We hypothesized that (1) compared to statistical algorithms, some machine learning models could offer a promising quantitative approach with high accuracy for potential invader prediction; (2) native range distribution size, origins and life-history traits co-determine an alien plant’s performance in the latter invasion stage. In this study, we used four machine learning models [classification and regression tree (CART), multivariate adaptive regression spline (MARS), random forest (RF) and multiple additive regression tree (MART)] and two traditional statistical algorithms [logistic regression (LR) and linear discriminant analysis (LDA)] to assess the relative importance of biogeographical and trait variables in the naturalized-invasion stage of 150 invasive and 87 non-invasive herb plants in China. Our results showed that good performance was the case for all predictive models (AUROC ranges from 0.68 to 0.87), which had overall mean performance value ranging from 0.66 to 0.82. Compared with traditional statistical algorithms, MART and RF models have a consistently higher accuracy, indicating that these two models could be used as alternative quantitative approaches for risk assessment. Additionally, both biogeographical (native range distribution size) and life-history traits (seed weight) were screened out by the models, suggesting their high correlation with plant invasiveness and important roles in risk assessment.

The National Environmental Respiratory Center (NERC) experiment in multi-pollutant air quality health research: IV. Vascular effects of repeated inhalation exposure to a mixture of five inorganic gases

An experiment was conducted to test the hypothesis that a mixture of five inorganic gases could reproduce certain central vascular effects of repeated inhalation exposure of apolipoprotein E-deficient mice to diesel or gasoline engine exhaust. The hypothesis resulted from preceding multiple additive regression tree (MART) analysis of a composition–concentration–response database of mice exposed by inhalation to the exhausts and other complex mixtures. The five gases were the predictors most important to MART models best fitting the vascular responses. Mice on high-fat diet were exposed 6 h/d, 7 d/week for 50 d to clean air or a mixture containing 30.6 ppm CO, 20.5 ppm NO, 1.4 ppm NO2, 0.5 ppm SO2, and 2.0 ppm NH3 in air. The gas concentrations were below the maxima in the preceding studies but in the range of those in exhaust exposure levels that caused significant effects. Five indicators of stress and pro-atherosclerotic responses were measured in aortic tissue. The exposure increased all five response indicators, with the magnitude of effect and statistical significance varying among the indicators and depending on inclusion or exclusion of an apparent outlying control. With the outlier excluded, three responses approximated predicted values and two fell below predictions. The results generally supported evidence that the five gases drove the effects of exhaust, and thus supported the potential of the MART approach for identifying putative causal components of complex mixtures.

The National Environmental Respiratory Center (NERC) experiment in multi-pollutant air quality health research: III. Components of diesel and gasoline engine exhausts, hardwood smoke and simulated downwind coal emissions driving non-cancer biological responses in rodents

An approach to identify causal components of complex air pollution mixtures was explored. Rats and mice were exposed by inhalation 6 h daily for 1 week or 6 months to dilutions of simulated downwind coal emissions, diesel and gasoline exhausts and wood smoke. Organ weights, hematology, serum chemistry, bronchoalveolar lavage, central vascular and respiratory allergic responses were measured. Multiple additive regression tree (MART) analysis of the combined database ranked 45 exposure (predictor) variables for importance to models best fitting 47 significant responses. Single-predictor concentration-response data were examined for evidence of single response functions across all exposure groups. Replication of the responses by the combined influences of the two most important predictors was tested. Statistical power was limited by inclusion of only four mixtures, albeit in multiple concentrations each and with particles removed for some groups. Results gave suggestive or strong evidence of causation of 19 of the 47 responses. The top two predictors of the 19 responses included only 12 organic and 6 inorganic species or classes. An increase in red blood cell count of rats by ammonia and pro-atherosclerotic vascular responses of mice by inorganic gases yielded the strongest evidence for causation and the best opportunity for confirmation. The former was a novel finding; the latter was consistent with other results. The results demonstrated the plausibility of identifying putative causal components of highly complex mixtures, given a database in which the ratios of the components are varied sufficiently and exposures and response measurements are conducted using a consistent protocol.

The National Environmental Respiratory Center (NERC) experiment in multi-pollutant air quality health research: I. Background, experimental strategy and critique

The National Environmental Respiratory Center Program was initiated as an experiment to explore strategies for identifying the components of complex air pollution mixtures that cause health effects associated statistically with air pollution. A strategy involving multivariate analysis of a composition-concentration-response database was adopted. A novel database was created by exposing rodents daily for up to six months to one of four combustion-related mixtures and measuring respiratory, cardiovascular and general toxicological responses after one week or six months of exposure. The mixtures included multiple concentrations of diesel and gasoline engine exhaust, hardwood smoke and simulated downwind coal combustion emissions. After reporting the biological effects of each mixture and comparing effects among them, 47 significant effects were selected for multiple additive regression tree analysis to identify putative causal components. Although the four mixtures provided a database marginally sufficient for the analysis, the results suggested the putative causes of 19 significant effects with acceptable confidence. This article describes and critiques the Program and its strategy. The integrated results are presented in two accompanying papers, and mixture-specific results were presented in preceding papers, which are cited. The experiment demonstrated the potential utility of the general approach and identified certain cause–effect relationships for confirmatory studies. A follow-up study provided support for causation by the components implicated for one of those relationships. The advantages and disadvantages of the Program’s management and funding strategies are discussed.

Optimizing the Prediction Accuracy of Concrete Compressive Strength Based on a Comparison of Data-Mining Techniques

This study attempts to optimize the prediction accuracy of the compressive strength of high-performance concrete (HPC) by comparing data-mining methods. Modeling the dynamics of HPC, which is a highly complex composite material, is extremely challenging. Concrete compressive strength is also a highly nonlinear function of ingredients. Several studies have independently shown that concrete strength is determined not only by the water-to-cement ratio but also by additive materials. The compressive strength of HPC is a function of all concrete content, including cement, fly ash, blast-furnace slag, water, superplasticizer, age, and coarse and fine aggregate. The quantitative analyses in this study were performed by using five different data-mining methods: two machine learning models (artificial neural networks and support vector machines), one statistical model (multiple regression), and two metaclassifier models (multiple additive regression trees and bagging regression trees). The methods were developed and tested against a data set derived from 17 concrete strength test laboratories. The cross-validation of unbiased estimates of the prediction models for performance comparison purposes indicated that multiple additive regression tree (MART) was superior in prediction accuracy, training time, and aversion to overfitting. Analytical results suggested that MART-based modeling is effective for predicting the compressive strength of varying HPC age.

A multiple additive regression tree analysis of three exposure measures during Hurricane Katrina

This paper analyses structural and personal exposure to Hurricane Katrina. Structural exposure is measured by flood height and building damage; personal exposure is measured by the locations of 911 calls made during the response. Using these variables, this paper characterises the geography of exposure and also demonstrates the utility of a robust analytical approach in understanding health-related challenges to disadvantaged populations during recovery. Analysis is conducted using a contemporary statistical approach, a multiple additive regression tree (MART), which displays considerable improvement over traditional regression analysis. By using MART, the percentage of improvement in R-squares over standard multiple linear regression ranges from about 62 to more than 100 per cent. The most revealing finding is the modelled verification that African Americans experienced disproportionate exposure in both structural and personal contexts. Given the impact of exposure to health outcomes, this finding has implications for understanding the long-term health challenges facing this population.

Soil carbon stocks, deforestation and land‐cover changes in the Western Ghats biodiversity hotspot (India)

AbstractHabitat loss and soil organic carbon (SOC) stock variations linked to land‐cover change were estimated over two decades in the most densely populated biodiversity hotspot in the world, in order to assess the possible influence of conservation practices on the protection of SOC. For a study area of 88 484 km2, 70% of which lie inside the Western Ghats Biodiversity Hotspot (WGBH), land‐cover maps for two dates (1977, 1999) were built from various data sources including remote sensing images and ecological forest maps. SOC stocks were calculated from climatic parameters, altitude, physiography, rock type, soil type and land‐cover, with a modelling approach used in predictive learning and based on Multiple Additive Regression Tree. The model was trained on 361 soil profiles data, and applied to estimate SOC stocks from predictor variables using a Geographical Information System (GIS). Comparison of 1977 and 1999 land‐cover maps showed 628 km2 of dense forests habitat loss (6%), corresponding to an annual deforestation rate of 0.44%. This was found consistent with other studies carried out in other parts of the WGBH, but not with FAO figures showing an increase in forest area. This could be explained by the different forest definitions used, based on ecological classification in the former, and on percentage tree cover in the latter. Unexpectedly, our results showed that despite ongoing deforestation, overall SOC stock was maintained (∼0.43 Pg). But a closer examination of spatial differences showed that soil carbon losses in deforested areas were compensated by sequestration elsewhere, mainly in recent plantations and newly irrigated croplands. This suggests that more carbon sequestration in soils could be achieved in the future through appropriate wasteland management. It is also expected that increasing concerns about biodiversity loss will favour more conservation and reinforce the already prevailing protective measures, thus further maintaining C stocks.

What Characteristics of Applicants to Emergency Medicine Residency Programs Predict Future Success as an Emergency Medicine Resident?

Program directors of emergency medicine (EM) residencies attempt to select candidates who will subsequently perform well as residents. This study was undertaken to identify characteristics available at the time of application to an EM residency that predict future success in residency. The EM faculty at the University of California San Diego (UCSD) completed a one-time confidential assessment of EM residents on performance in residency at the time of graduation. The faculty member compared the graduate with all residents (both EM and non-EM) with whom the faculty member had previously worked using the five-point scale: > or =90th percentile, 70th-89th percentile, 50th-69th percentile, 30th-49th percentile, or < 30th percentile. Descriptive statistics, ordinal logistic regression (OLR), classification and regression tree (CART) analysis, and multiple additive regression tree (MART) analysis were used to find predictors for each of the outcome variables. Fifty-four graduates were evaluated. The medical school attended (MSA) was the strongest predictor of overall performance in residency in all regression models. OLR showed that MSA and "distinctive factors" (being a championship athlete, medical school officer, etc.) were significant predictors and may deserve greater weighting in the selection process. The most robust MART model demonstrated that MSA, dean's letter of recommendation, and distinctive factors had the most impact on overall performance in an EM residency. Using regression modeling, it may be possible to predict future resident performance from characteristics contained in residency applications. Applicants from top-tier medical schools and those with distinctive talents were more successful in the UCSD EM residency.

Multiple additive regression trees with application in epidemiology.

Predicting future outcomes based on knowledge obtained from past observational data is a common application in a wide variety of areas of scientific research. In the present paper, prediction will be focused on various grades of cervical preneoplasia and neoplasia. Statistical tools used for prediction should of course possess predictive accuracy, and preferably meet secondary requirements such as speed, ease of use, and interpretability of the resulting predictive model. A new automated procedure based on an extension (called 'boosting') of regression and classification tree (CART) models is described. The resulting tool is a fast 'off-the-shelf' procedure for classification and regression that is competitive in accuracy with more customized approaches, while being fairly automatic to use (little tuning), and highly robust especially when applied to less than clean data. Additional tools are presented for interpreting and visualizing the results of such multiple additive regression tree (MART) models.