All-subsets Regression Research Articles

Integrated CT Radiomics Features Could Enhance the Efficacy of 18F-FET PET for Non-Invasive Isocitrate Dehydrogenase Genotype Prediction in Adult Untreated Gliomas: A Retrospective Cohort Study.

PurposeWe aimed to investigate the predictive models based on O-[2-(18F)fluoroethyl]-l-tyrosine positron emission tomography/computed tomography (18F-FET PET/CT) radiomics features for the isocitrate dehydrogenase (IDH) genotype identification in adult gliomas.MethodsFifty-eight consecutive pathologically confirmed adult glioma patients with pretreatment 18F-FET PET/CT were retrospectively enrolled. One hundred and five radiomics features were extracted for analysis in each modality. Three independent radiomics models (PET-Rad Model, CT-Rad Model and PET/CT-Rad Model) predicting IDH mutation status were generated using the least absolute shrinkage and selection operator (LASSO) regression analysis based on machine learning algorithms. All-subsets regression and cross validation were applied for the filter and calibration of the predictive radiomics models. Besides, semi-quantitative parameters including maximum, peak and mean tumor to background ratio (TBRmax, TBRpeak, TBRmean), standard deviation of glioma lesion standardized uptake value (SUVSD), metabolic tumor volume (MTV) and total lesion tracer uptake (TLU) were obtained and filtered for the simple model construction with clinical feature of brain midline involvement status. The area under the receiver operating characteristic curve (AUC) was applied for the evaluation of the predictive models.ResultsThe AUC of the simple predictive model consists of semi-quantitative parameter SUVSD and dichotomized brain midline involvement status was 0.786 (95% CI 0.659-0.883). The AUC of PET-Rad Model building with three 18F-FET PET radiomics parameters was 0.812 (95% CI 0.688-0.902). The AUC of CT-Rad Model building with three co-registered CT radiomics parameters was 0.883 (95% CI 0.771-0.952). While the AUC of the combined 18F-FET PET/CT-Rad Model building with three CT and one PET radiomics features was 0.912 (95% CI 0.808-0.970). DeLong test results indicated the PET/CT-Rad Model outperformed the PET-Rad Model (p = 0.048) and simple predictive model (p = 0.034). Further combination of the PET/CT-Rad Model with the clinical feature of dichotomized tumor location status could slightly enhance the AUC to 0.917 (95% CI 0.814-0.973).ConclusionThe predictive model combining 18F-FET PET and integrated CT radiomics features could significantly enhance and well balance the non-invasive IDH genotype prediction in untreated gliomas, which is important in clinical decision making for personalized treatment.

A nomogram to predict postoperative infection for older hip fracture patients.

Postoperative infection is one of the most common postoperative complications in hip fracture surgery. It is related with increased morbidity and mortality. This study aimed at developing a nomogram to predict the individual probability of postoperative infection to facilitate perioperative decision-making. In this retrospective study, we included all patients over 65years old admitted for hip fracture in West China Hospital of Sichuan University from 1 January 2015 to 31 December 2019. Univariate and multivariate logistic regression analyses were used to identify significant predictors. We used all-subsets regression to screen an optimal model, and visualized the model through drawing nomogram. To evaluate the model performance, we applied receiver operating characteristic curve and calibration curve. We enrolled 677 older patients. 136 (20.1%) patients developed postoperative infection during hospitalization. Variables retained in the final model were albumin [odds ratio (OR) 0.90, 95% confidence interval (CI) 0.84-0.96], cholesterol (OR 1.49, 95% CI 1.04-2.15), blood phosphorus (OR 0.16, 95% CI 0.05-0.48), high-density lipoprotein (OR 0.42, 95% CI 0.19-0.89), surgery type (OR 2.27, 95% CI 1.35-3.90), smoking (OR 1.95, 95% CI 1.02-3.66), American Society of Anesthesiologists classification [class III (OR 1.02, 95% CI 0.55-1.93); class IV (OR 1.93, 95% CI 0.76-4.82)], and chronic pulmonary disease (OR 2.16, 95% CI 1.25-3.68). The C-index of the nomogram was 0.752 (95% CI 0.697-0.806). Calibration curve showed good agreement between predicted value and observed outcome. In the validation group, our nomogram showed an area under the receiver operating characteristic curve of 0.723 (95% CI 0.639-0.807). Our nomogram showed good discrimination ability in predicting individual probability of postoperative infection among older patients with hip fracture surgery. The nomogram could help clinicians identify patients at high risk of postoperative infection before surgery.

A real-time occupancy detection system for unoccupied, normally and abnormally occupied situation discrimination via sensor array and cloud platform in indoor environment

It is significant to detect the occupancy of indoor environment from the view of saving energy. In addition, occupied situation detection can help control the density of people in room to reduce the risk of disease transmission. In preliminary, we trained nine algorithm models on the existing occupancy dataset to select the optimum algorithm. Then, all-subsets regression model is used for feature selection (sensor contribution evaluation) to optimize the size of sensor array. As a result, the voting based weighted extreme learning machine (WV-ELM) model achieved the highest prediction accuracy and the combination of light and CO2 sensors could realize a satisfied classification result. Finally, a real-time occupancy detection system based on the sensor array and cloud platform was proposed. The system used the indoor environmental data collected by the sensor array. WV-ELM model was combined with the proposed system to detect the real-time occupied situation in the indoor environment. To verify their efficiency, the system was implemented in a laboratory to collect occupancy data for one week. According to the actual test results, the proposed system realized a detection accuracy of 97.32% with running time less than 30 s.

Internal transcribed spacer sequencing and metatranscriptomics analysis reveal the fungal community composition, diversity related environment variables and roles during serofluid dish fermentation

Serofluid dish is a popular fermented food in northwest China that contains abundant microorganisms. Advances of next-generation sequencing facilitate the elucidation of bacterial community composition, yet roles of fungi in serofluid dish need further exploration. Here, internal transcribed spacer (ITS) sequencing and metatranscriptomics were conducted and results showed that the dominant fungal communities belong to Candida and Saccharomyces. Additionally, fungal diversity, including phylogenetic diversity and species richness, was largely correlated with pH, the total organic carbon (TOC) and spatial distance. Multivariate regression tree and all-subsets regression analysis also identified them as strong predictors. Further gene annotations revealed that fungi exhibited higher participation in the carbohydrate metabolism, amino acid metabolism, translation and energy metabolism pathways, including sucrose and maltose metabolism and the production of acetate, fumarate and ethanol. Collectively, fungi in serofluid dish possess rich species diversity, and are affected by both environmental variation and geographical distance. In the fermentation process, fungi predominantly participated in the carbohydrate metabolism and genetic information processing.

Waterlogging Resistance Evaluation Index and Photosynthesis Characteristics Selection: Using Machine Learning Methods to Judge Poplar’s Waterlogging Resistance

Flood disasters are the major natural disaster that affects the growth of agriculture and forestry crops. Due to rapid growth and strong waterlogging resistance characteristics, many studies have explained the waterlogging resistance mechanism of poplar from different perspectives. However, there is no accurate method to define the evaluation index of waterlogging resistance. In addition, there is also a lack of research on predicting the waterlogging resistance of poplars. Based on the changes of poplar biomass and seedling height, the evaluation index of poplar resistance to waterlogging was well determined, and the characteristics of photosynthesis were used to predict the waterlogging resistance of poplars. First, four methods of hierarchical clustering, lasso, stepwise regression and all-subsets regression were used to extract the photosynthesis characteristics. After that, the support vector regression model of poplar resistance to waterlogging was established by using the characteristic parameters of photosynthesis. Finally, the results show that the SVR model based on Stepwise regression and Lasso method has high precision. On the test set, the coefficient of determination (R2) was 0.8581 and 0.8492, the mean square error (MSE) was 0.0104 and 0.0341, and the mean relative error (MRE) was 9.78% and 9.85%, respectively. Therefore, using the characteristic parameters of photosynthesis to predict the waterlogging resistance of poplars is feasible.

Model Selection Reveals the Butyrate-Producing Gut Bacterium Coprococcus eutactus as Predictor for Language Development in 3-Year-Old Rural Ugandan Children.

IntroductionThe metabolic activity of the gut microbiota plays a pivotal role in the gut-brain axis through the effects of bacterial metabolites on brain function and development. In this study we investigated the association of gut microbiota composition with language development of 3-year-old rural Ugandan children.MethodsWe studied the language ability in 139 children of 36 months in our controlled maternal education intervention trial to stimulate children’s growth and development. The dataset includes 1170 potential predictors, including anthropometric and cognitive parameters at 24 months, 542 composition parameters of the children’s gut microbiota at 24 months and 621 of these parameters at 36 months. We applied a novel computationally efficient version of the all-subsets regression methodology and identified predictors of language ability of 36-months-old children scored according to the Bayley Scales of Infant and Toddler Development (BSID-III).ResultsThe best three-term model, selected from more than 266 million models, includes the predictors Coprococcus eutactus at 24 months of age, Bifidobacterium at 36 months of age, and language development at 24 months. The top 20 four-term models, selected from more than 77 billion models, consistently include C. eutactus abundance at 24 months, while 14 of these models include the other two predictors as well. Mann–Whitney U tests suggest that the abundance of gut bacteria in language non-impaired children (n = 78) differs from that in language impaired children (n = 61). While anaerobic butyrate-producers, including C. eutactus, Faecalibacterium prausnitzii, Holdemanella biformis, Roseburia hominis are less abundant, facultative anaerobic bacteria, including Granulicatella elegans, Escherichia/Shigella and Campylobacter coli, are more abundant in language impaired children. The overall predominance of oxygen tolerant species in the gut microbiota was slightly higher in the language impaired group than in the non-impaired group (P = 0.09).ConclusionApplication of the all-subsets regression methodology to microbiota data established a correlation between the relative abundance of the anaerobic butyrate-producing gut bacterium C. eutactus and language development in Ugandan children. We propose that the gut redox potential and the overall bacterial butyrate-producing capacity in the gut are important factors for language development.

Identifying the dominant driving factors of heat waves in the North China Plain

The driving factors of heat waves (HW) have been analyzed widely, but the relative importance of these factors is still poorly understood. In this study, the temporal and spatial variations of HW in the North China Plain (NCP) were investigated using meteorological datasets during 1961–2014. The relationship between HW and several potential driving factors related to climate change (air temperature, precipitation and solar radiation), anthropogenic activity (urbanization, irrigation and aerosol optical depth (AOD)), land-surface interaction (soil moisture, drought and land surface albedo) and atmospheric circulation were assessed. The relative importance of these factors was revealed based on all-subsets regression and hierarchical partitioning (HP) analysis. Results showed that HW in the NCP became increasingly intense, frequent, durable, and emerged earlier and terminated later, with a significant increased slope in intensity, frequency duration and timing. Spatially, greater upward trends were observed in the northern NCP. Climate warming, urbanization, irrigation, AOD, land surface albedo and Atlantic Multidecadal Oscillation were significantly correlated with HW. Based on HP analysis, the most important driving factor was climate warming, contributing approximately 40% to HW variation. The cumulative contribution rate of anthropogenic activity to the HW frequency, duration and amplitude reached nearly 35%. The AMO and land surface albedo explained approximately 15% and 20% of the HW variation, respectively. This study helps to understand the characteristics of regional extreme heat events and provides a scientific basis for developing HW alleviation strategies.

Adult Body Height and Cardiometabolic Disease Risk: The China National Health Survey in Shaanxi.

ObjectivesBased on data from the China National Health Survey, we aimed to examine the association between body height and cardiometabolic disease (CMD) in a large adult population from Shaanxi province, and further to test whether this association was hinged upon other population characteristics.MethodsThis population-based study was conducted in 2014 in Shaanxi Province, China. Utilizing a multi-stage stratified cluster sampling method, total 5,905 adults with complete data were eligible for analysis, and 1,151 (19.5%) of them had CMD. Of 1,151 CMD patients, 895 (15.1%) had one disorder and 256 (4.4%) had ≥2 disorders.ResultsUsing the bi-directional stepwise method and all-subsets regression, five factors—age, body mass index, family histories of CMD, exercise, and height—constituted the optimal model when predicting CMD risk. Restricted cubic spline regression showed a reduced tendency towards CMD with the increase of body height, with per 10 cm increment in body height corresponding to 14% reduced risk. Ordinal Logistic regression supported the contribution of body height on both continuous and categorical scales to CMD risk before and after adjustment, yet this contribution was significantly confounded by exercise and education, especially by exercise, which can explain 65.4% of total impact. For example, short stature was associated with an increased risk of CMD after multivariable adjustment not including exercise and education (odds ratio, 95% confidence interval, P: 1.42, 1.21 to 1.66, <0.001), and tall stature was associated with a reduced risk (0.77, 0.64 to 0.92, 0.003).ConclusionsOur findings indicate short stature was a risk factor, yet tall stature was a protective factor for CMD in Chinese. Notably, the prediction of short and tall stature for CMD may be mediate in part by exercise.

Comparison of multi-source satellite data for quantifying water quality parameters in a mining environment

Population growth and associated anthropogenic activities such as mining are posing a major threat to water resources. Remote sensing provides valuable information for assessing and monitoring water quality and quantity, and thus supports sustainable water management. Several studies have demonstrated the utility of various remote sensing techniques for quantifying water quality parameters, however, more needs to be done to exploit the growing number of satellite sensors available to the public. The main objective of this study was to evaluate the performances of individual and combined bands of Landsat 8 OLI, Sentinel-2 MSI, ASTER and SPOT 6 data as predictors of water quality parameters of open water bodies in a mining area. We applied the all-subsets regression approach to explore all possible candidate models from which selection was made based on Akaike’s Information Criterion (AIC), coefficient of determination (adjR2) and Root-Mean-Squared-Error (RMSE). Two secondary objectives that make use of the products of the all-subsets regression approach were added in the study. This included comparison of a parsimonious model containing few predictors with the model that was ranked the best but containing more predictors; and comparison of models consisting of similar bands across sensors per water quality parameter. In general, all sensors yielded promising estimation accuracies, particularly for SAR, permanent hardness and cations (RMSE < 50% of observed means). Sentinel-2 bands produced the best estimations followed by Landsat 8 bands for most water quality parameters, chiefly due to the greater number of bands these sensors possess compared to ASTER and SPOT 6 sensors. All sensors provided alternative models that used fewer predicting bands than the best-ranked models (lowest AIC) and still showed comparable accuracies with the lowest AIC models for most variables (Pearson’s correlation, r up to 0.9). Sentinel-2, in particular, had the most accurate, parsimonious models compared to the other sensors. Pairwise cross-sensor comparison of estimation models using similar bands revealed the best matches when a model consisted of blue band only (r > 0.95) for most parameters while the similarity decreased as the number of predicting bands increased. Furthermore, the best similarities were obtained between Sentinel-2 and Landsat 8 sensors compared to the other pairs, indicating the potential of replacing one in the absence of the other. The findings of this study showed the advantage of commonly used remotely-sensed data for estimating, and by extension monitoring, water quality parameters in open water bodies. In addition, the results revealed that few, specific bands carry information relevant to characterise most of the water quality parameters considered in the study. We recommend similar studies for different seasons, considering the need to monitor water quality throughout the year.

Growth responses and physiological and biochemical changes in five ornamental plants grown in urban lead-contaminated soils.

An increasing concentration of lead (Pb) in urban contaminated soil due to anthropogenic activities has been a global issue threatening human health. The use of urban ornamental plants as phytoremediation of Pb-contaminated soil is a new choice. In the present experiment, the physiological and biochemical response of five ornamental plants to increase in concentrations of C4H6O4Pb·H2O in the soil were measured to investigate these plans' Pb tolerance strategies and abilities. Our results showed that Pb stress significantly inhibited the growth and the biomass of all the plants. The root activity (RA), net photosynthetic rate (P n), and chlorophyll (Chl) content in Pb-stressed leaves were significantly decreased, whereas the leaf proline (Pro), soluble sugar (SS), and membrane stability index (MSI) were remarkable increased compared with those in the control group. By application of all-subsets regression and linear regression, the reduction in photosynthetic capacity in the five plants is mainly due to the decrease in the leaf Chl content caused by Pb stress. The bioconcentration factor (BCF) in Canna generalis was greater than 1, while in the other plants were lower than 1, suggesting that Canna generalis had the highest Pb accumulation ability. The translocation factor (TF) in all the plants were lower than 1, suggesting that Pb preferentially accumulated in the external part of roots. By calculating the comprehensive evaluation value (CEV), Iris germanica L. was found to be the most sensitive species, and Canna generalis was the most tolerant species, to Pb stress among the five ornamental plants.

Quantitative analysis of the interannual variation in the seasonal water cooling island (WCI) effect for urban areas

Water body, as water cool islands (WCIs), is an effective factor of mitigating the urban heat island (UHI) effects through the evaporation and absorption of solar shortwave radiance. In this study, a combination of the range, amplitude, efficiency and capacity of the WCI effect (WCI_R, WCI_A, WCI_E and WCI_C as descriptors) within water buffers, is proposed to evaluate the interannual variation of the seasonal WCI effects within the fifth loop of Beijing from 2000 to 2018, further construct optimal regression models by seven impact factors with the four WCI descriptors through all-subset regression, as well as calculate the independent contribution rate (CR) of these factors by hierarchical partitioning (HP) analysis. The results indicate that (1) In terms of the interannual variation of WCI effects, the most significant trends that rise over fall produced by four descriptors generate in summer after 2009, while relatively stable states exist in spring and winter. (2) As for the impact factors, the water temperature (WT) and the percentage of vegetation (PV) contribute most for WCI_R, WCI_A and WCI_E in reaction to the interannual average, while the water temperature capacity (WTC) maintains significant impact on WCI_C from spring to autumn over 18 years but shares power with water temperature (WA) and WT in winter. (3) Considering the factor changes yearly, the dominance relationships of external ones, mainly the percentage of impervious surface (PI) and PV, gradually surpass that of WA and WTC in four seasons especially after 2009. On this basis, urban designers could propose suitable plans to better exert the cooling effect of water bodies, thereby reducing the UHI effect.

LmSubsets: Exact Variable-Subset Selection in Linear Regression for R

An R package for computing the all-subsets regression problem is presented. The proposed algorithms are based on computational strategies recently developed. A novel algorithm for the best-subset regression problem selects subset models based on a predetermined criterion. The package user can choose from exact and from approximation algorithms. The core of the package is written in C++ and provides an efficient implementation of all the underlying numerical computations. A case study and benchmark results illustrate the usage and the computational efficiency of the package.

Surface Heterogeneity-Involved Estimation of Sample Size for Accuracy Assessment of Land Cover Product from Satellite Imagery.

Sample size estimation is a key issue for validating land cover products derived from satellite images. Based on the fact that present sample size estimation methods account for the characteristics of the Earth’s subsurface, this study developed a model for estimating sample size by considering the scale effect and surface heterogeneity. First, we introduced a watershed with different areas to indicate the scale effect on the sample size. Then, by employing an all-subsets regression feature selection method, three landscape indicators describing the aggregation and diversity of the land cover patches were selected (from 14 indicators) as the main factors for indicating the surface heterogeneity. Finally, we developed a multi-level linear model for sample size estimation using explanatory variables, including the estimated sample size (n) calculated from the traditional statistical model, size of the test region, and three landscape indicators. As reference data for developing this model, we employed a case study in the Jiangxi Province using a 30 m spatial resolution global land cover product (Globeland30) from 2010 as a classified map, and national 30 m land use/cover change (LUCC) data from 2010 in China. The results showed that the adjusted square coefficient of R2 is 0.79, indicating that the joint explanatory ability of all predictive variables in the model to the sample size is 79%. This means that the predictability of this model is at a good level. By comparing the sample size NS obtained by the developed multi-level linear model and n as calculated from the statistics model, we find that NS is much smaller than n, which mainly contributes to the concerns regarding surface heterogeneity in this study. The validity of the established model is tested and is proven as effective in the Anhui Province. This indicates that the estimated sample size from considering the scale effect and spatial heterogeneity in this study achieved the same accuracy as that calculated from a probability statistical model, while simultaneously saving more time, labour, and money in the accuracy assessment of a land cover dataset.

A brief version of the questionnaire of olfactory disorders in patients with chronic rhinosinusitis.

The Questionnaire of Olfactory Disorders-Negative Statements (QOD-NS) is a valuable instrument in the measurement of olfactory-specific quality of life (QOL). In the clinical setting, patients can be overwhelmed with the time required to complete questionnaires. Our objective was to develop a brief version of the QOD-NS to streamline clinical care and research. QOD-NS scores from 221 subjects were used to determine which subset of the 17 QOD-NS questions best correlated with total and subdomain QOD-NS scores. An initial pool of 11 questions was made by removing items with ρ < 0.80 to their respective subdomain scores. Next, 500 bootstrapped samples were taken. On each sampe, an all-subsets regression was performed with total QOD-NS scores and QOD-NS subdomain scores as the outcomes. From this, our "top" and "bottom" 10 subsets were identified based on mean r2 value, representation in bootstrap analysis, and number of items. All of our top subsets had excellent correlation with total and subdomain QOD-NS scores (mean r2 > 0.90). Our top choice has 7 total questions, is representative of all subdomains, has a mean r2 = 0.92, and was represented in 323 of our 500 bootstrapped samples. The worst-performing subset has 5 items, mean r2 = 0.81, and was represented in only 1 bootstrapped sample. Using less than half of the questions in the QOD-NS, excellent correlations with both total and domain-specific scores are achieved. A brief version of the QOD-NS may prove useful in future clinical and research settings.

Exploring cortical predictors of clinical response to electroconvulsive therapy in major depression.

Electroconvulsive therapy (ECT) is a rapid and highly effective treatment option for treatment-resistant major depressive disorder (TRD). The neural mechanisms underlying such beneficial effects are poorly understood. Exploring associations between changes of brain structure and clinical response is crucial for understanding ECT mechanisms of action and relevant for the validation of potential biomarkers that can facilitate the prediction of ECT efficacy. The aim of this explorative study was to identify cortical predictors of clinical response in TRD patients treated with ECT. We longitudinally investigated 12 TRD patients before and after ECT. Twelve matched healthy controls were studied cross sectionally. Demographical, clinical, and structural magnetic resonance imaging data at 3T and multiple cortical markers derived from surface-based morphometry (SBM) analyses were considered. Multiple regression models were computed to identify predictors of clinical response to ECT, as reflected by Hamilton Depression Rating Scale (HAMD) score changes. Symptom severity differences pre-post-ECT were predicted by models including demographic data, clinical data and SBM of frontal, cingulate, and entorhinal structures. Using all-subsets regression, a model comprising HAMD score at baseline and cortical thickness of the left rostral anterior cingulate gyrus explained most variance in the data (multiple R2 = 0.82). The data suggest that SBM provides powerful measures for identifying biomarkers for ECT response in TRD. Rostral anterior cingulate thickness and HAMD score at baseline showed the greatest predictive power of clinical response, in contrast to cortical complexity, cortical gyrification, or demographical data.

Seasonal contrast of the dominant factors for spatial distribution of land surface temperature in urban areas

Urban heat island (UHI) has become an urban eco-environmental problem globally. Land surface temperature (LST) is widely used to quantify UHI. This study used Shenzhen, a southern coastal city in China, as an example to explore the relationship between spatial variation of LST in different seasons and the influencing factors in five dimensions, integrating the methods of ordinary least-squares regression, stepwise regression, all-subsets regression, and hierarchical partitioning analysis. The results showed that the most important factor affecting spatial heterogeneity of LST in summer was the normalized difference build-up index (53.62%, for contributing rate), whereas in the transition season the most important factor was the normalized difference vegetation index (NDVI) (47.84%). In winter the construction land percentage and NDVI (26.84% and 25.56%, respectively) were the most influential. Artificial surface and green space had a dominant effect on LST spatial differentiation. Landscape configuration and diversity were not the dominant influencing factors in summer or in the transition season. Furthermore, the independent contribution rate of the Shannon diversity index (SHDI) reached 8.79% in the transition season, while in winter, the independent contribution rates of SHDI and the landscape shape index were 8.52% and 3.45%, respectively. The influence of landscape diversity and configuration factors tended to increase as LST reduced, while the contribution rate of the important factors such as artificial surface and green space decreased significantly. These relationships indicate that the influence of landscape configuration and diversity factors on LST is relatively weak, and can be easily concealed by the influence of landscape components, especially when the spatial variation of LST is not strong. These findings can help to develop UHI adaptation strategies based on local conditions.

Predictive QSAR Models for the Toxicity of Disinfection Byproducts.

Several hundred disinfection byproducts (DBPs) in drinking water have been identified, and are known to have potentially adverse health effects. There are toxicological data gaps for most DBPs, and the predictive method may provide an effective way to address this. The development of an in-silico model of toxicology endpoints of DBPs is rarely studied. The main aim of the present study is to develop predictive quantitative structure–activity relationship (QSAR) models for the reactive toxicities of 50 DBPs in the five bioassays of X-Microtox, GSH+, GSH−, DNA+ and DNA−. All-subset regression was used to select the optimal descriptors, and multiple linear-regression models were built. The developed QSAR models for five endpoints satisfied the internal and external validation criteria: coefficient of determination (R2) > 0.7, explained variance in leave-one-out prediction (Q2LOO) and in leave-many-out prediction (Q2LMO) > 0.6, variance explained in external prediction (Q2F1, Q2F2, and Q2F3) > 0.7, and concordance correlation coefficient (CCC) > 0.85. The application domains and the meaning of the selective descriptors for the QSAR models were discussed. The obtained QSAR models can be used in predicting the toxicities of the 50 DBPs.

Comprehensive Clusterwise Linear Regression for Pavement Management Systems

A comprehensive mathematical program was formulated to determine simultaneously (1) an optimum number of pavement clusters, (2) cluster memberships of pavement samples, (3) cluster-specific significant explanatory variables, and (4) estimated regression coefficients for pavement performance models (PPMs). Simulated annealing coupled with all-subset regression was proposed to solve the mathematical programming. The proposed algorithm was capable of identifying and addressing potential multicollinearity issues. All possible combinations of the explanatory variables were examined to select the best model that provided a balance among (1) the number of PPMs; (2) the number of explanatory variables; (3) the resources required to develop, maintain, and use these models; and (4) the explanatory power. For the data set used in this research, six-cluster models were determined as part of the optimum solution. The predictive capabilities of the resultant models were investigated, and results showed that the models provided few prediction errors without any overfitting issues.

Evaluation of MODIS Land Surface Temperature Data to Estimate Near-Surface Air Temperature in Northeast China

Air temperature (Tair) near the ground surface is a fundamental descriptor of terrestrial environment conditions and one of the most widely used climatic variables in global change studies. The main objective of this study was to explore the possibility of retrieving high-resolution Tair from the Moderate Resolution Imaging Spectroradiometer (MODIS) land surface temperature (LST) products, covering complex terrain in Northeast China. The All Subsets Regression (ASR) method was adopted to select the predictors and build optimal multiple linear regression models for estimating maximum (Tmax), minimum (Tmin), and mean (Tmean) air temperatures. The relative importance of predictors in these models was evaluated via the Standardized Regression Coefficients (SRCs) method. The results indicated that the optimal models could estimate the Tmax, Tmin, and Tmean with relatively high accuracies (Model Efficiency ≥ 0.90). Both LST and day length (DL) predictors were important in estimating Tmax (SRCs: daytime LST = 0.53, DL = 0.35), Tmin (SRCs: nighttime LST = 0.74, DL = 0.23), and Tmean (SRCs: nighttime LST = 0.72, DL = 0.28). Models predicting Tmin and Tmean had better performance than the one predicting Tmax. Nighttime LST was better at predicting Tmin and Tmean than daytime LST data at predicting Tmax. Land covers had noticeable influences on estimating Tair, and even seasonal vegetation greening could result in temporal variations of model performance. Air temperature could be accurately estimated using remote sensing, but the model performance was varied across different spatial and temporal scales. More predictors should be incorporated for the purpose of improving the estimation of near surface Tair from the MODIS LST production.

Effective Design-Based Model Selection for Definitive Screening Designs

ABSTRACTSince their introduction by Jones and Nachtsheim in 2011, definitive screening designs (DSDs) have seen application in fields as diverse as bio-manufacturing, green energy production, and laser etching. One barrier to their routine adoption for screening is due to the difficulties practitioners experience in model selection when both main effects and second-order effects are active. Jones and Nachtsheim showed that for six or more factors, DSDs project to designs in any three factors that can fit a full quadratic model. In addition, they showed that DSDs have high power for detecting all the main effects as well as one two-factor interaction or one quadratic effect as long as the true effects are much larger than the error standard deviation. However, simulation studies of model selection strategies applied to DSDs can disappoint by failing to identify the correct set of active second-order effects when there are more than a few such effects. Standard model selection strategies such as stepwise regression, all-subsets regression, and the Dantzig selector are general tools that do not make use of any structural information about the design. It seems reasonable that a modeling approach that makes use of the known structure of a designed experiment could perform better than more general purpose strategies. This article shows how to take advantage of the special structure of the DSD to obtain the most clear-cut analytical results possible.