Residual Standard Deviation Research Articles

An automatically progressed computer-controlled simulated digestion system to predict digestible and metabolizable energy of unconventional plant protein meals for growing pigs.

The objective of this experiment was to develop a new computer-controlled simulated digestion system to predict the digestible energy (DE) and metabolizable energy (ME) of unconventional plant protein meals for growing pigs. Nine meals tested included 1 source of rapeseed meal, 4 sources of cottonseed meal, 2 sources of sunflower meal, and 2 sources of peanut meal. Twenty growing pigs (Duroc × [Landrace × Large White]) with an initial body weight (BW) of 41.7 ± 2.6 kg were allotted to a replicated 10 × 3 incomplete Latin square design to determine the DE and ME of 1 basal diet and 9 experimental diets formulated with 9 unconventional plant protein meals. The DE and ME values of unconventional plant protein meals were calculated by the difference method. The in vitro digestible energy (IVDE) of 1 basal diet, 9 experimental diets, and 9 unconventional plant protein meals were determined with 5 replicates of each sample in a complete randomized arrangement. The IVDE/DE or IVDE/ME ranged from 0.96 to 0.98 or 1.00 to 1.01, and the correlation coefficient between IVDE and DE or ME was 0.97 or 0.98 in 10 experimental diets. Accordingly, the IVDE/DE or IVDE/ME ranged from 0.86 to 1.05 or 0.96 to 1.20, and the correlation coefficient between IVDE and DE or ME was 0.92 or 0.91 in 9 unconventional plant protein meals. The coefficient of variation (CV) of IVDE was less than that of DE and ME in the experimental diets (0.43%, 0.80%, and 0.97% for CV of IVDE, DE and ME, respectively) and unconventional plant protein meals (0.92%, 4.84%, and 6.33% for CV of IVDE, DE and ME, respectively). The regression equations to predict DE from IVDE in 10 experimental diets and 9 unconventional plant protein meals were DE = 0.8851 × IVDE +539 (R2 = 0.9411, residual standard deviation [RSD] = 23 kcal/kg DM, P < 0.01) and DE = 0.9880 × IVDE + 166 (R2 = 0.8428, RSD = 182 kcal/kg DM, P < 0.01), respectively. There was no statistical difference in the slopes (P = 0.82) or intercepts (P = 1.00) of these 2 equations. Thus, 10 diets and 9 unconventional plant protein meals were pooled to establish the regression equation of DE on IVDE as: DE = 0.9813 × IVDE +187 (R2 = 0.9120, RSD = 118 kcal/kg DM, P < 0.01). The regression equations to predict ME from IVDE in 10 experimental diets and 9 unconventional plant protein meals were ME = 0.9559 × IVDE +146 (R2 = 0.9697, RSD = 18 kcal/kg DM, P < 0.01) and ME = 0.9388 × IVDE + 3 (R2 = 0.8282, RSD = 182 kcal/kg DM, P < 0.01), respectively. There was no statistical difference in slopes (P = 0.97) but significant difference between the intercepts (P = 0.02) of these 2 equations. Our results indicate IVDE has similar response to the DE but different response to the ME in 10 experimental diets and 9 unconventional plant protein meals. Therefore, IVDE is more suitable to predict DE than ME of diets and unconventional plant protein meals for growing pigs.

Application of generalized dispersion theory to vortex chromatography

Acoustically induced secondary flows are applied to enhance lateral mass transfer beyond the relatively slow diffusion. This has the goal to reduce convective axial dispersion and the resulting band broadening which, in turn, limits the performance of column chromatography. Traditional approaches based on Taylor-Aris model are limited to one-dimensional rectilinear (unidirectional) tube- or channel-flows. We therefore apply the generalized dispersion theory (GDT) allowing for prediction of the dependence of potentially improved performance on the characteristics of the induced secondary flow, channel geometry and solute properties as well as providing qualitative physical insight into the role of lateral flows. Results corroborate agreement with our experimental observations (residual standard deviation, Sres = 3.88) and demonstrate the advantage of applying GDT relative to 3D time-dependent simulations.

Accuracy of Heart Rate Measurements of Three Smartwatch Models in Dogs

This study was aimed to evaluate the accuracy of heart rate (HR) measurements using 3 smartwatch models designed for human use when applied to dogs. A group of 15 mixed breed dogs, ages ranging from 2 to 3 years, weighing between 20 and 30 kg were used. The Garmin Fenix 5X plus (GF5Xp), Samsung Gear S3 (SGS3), and Polygold A-6 (PDA6), were randomly placed on the tibia of dogs to collect the HR recordings every 5 minutes for 1 hour. A veterinary patient monitor (VPM) was used as the reference method. A Bland-Altman plot was applied to determine the difference between smartwatches. The receiver operating characteristic curves were developed analysis for sensitivity and specificity. The HR measured by VPM ranged from 65 to 200 bpm (mean, 149.3 ± 28.1 bpm; median, 154 bpm). Among the smartwatches, the lowest (44 bpm) HR was measured by PDA6, while the highest (201 bpm) was measured by SGS3. Among the smartwatches; GF5Xp and SGS3 had slopes equal to 1 and residual standard deviation closer to 0, in agreement with VPM. At the cut-off HR (160 bpm), SGS3 had the highest sensitivity (97.4%), whereas PDA6 had the lowest sensitivity (68.0%). The specificity of GF5Xp, SGS3, and PDA6 was 97.1%, 98.0%, and 98.0%, respectively. The SGS3 and GF5Xp can be ideal tools for veterinarians to monitor the HR of dogs without requiring an additional device.

Astrometric Calibration for All-sky Camera with Machine Learning

The night images obtained with an all-sky camera can provide spatial and time sampling, which can be used for measurement cloud coverage measurement and meteor monitoring. The astrometric calibration of an all-sky camera is necessary because of strong field distortions. We use machine learning to complete the calibration of an all-sky camera. In order to prepare the data sets needed for machine learning, a particle swarm optimization algorithm is used to determine the parameters of the method proposed by Borovicka in 1995. Machine learning can transform plate coordinates to celestial coordinates and transform celestial coordinates to plate coordinates. The actual test shows that the standard deviation of residuals is of the order of 1′ for the transformation from plate coordinates to celestial coordinates and the standard deviation of residuals is of the order of 0.3 px for the transformation from celestial coordinates to the plate coordinates.

Abstract P011: Blood Pressure Postural Changes Variability Is Associated With Cognitive Decline: The S.ages Cohort

Introduction: Over the past decade, higher blood pressure (BP) variability has been associated with cognitive decline (CD). To date, only one study investigated the specific impact of visit-to-visit BP postural changes variability on dementia but its association with cognition is still unknown. We aimed to investigate the association between BP postural changes variability and cognitive function in noninstitutionalized subjects aged ≥ 65 years. Methods: The S.AGES (Sujets Å;gés) cohort followed for three years noninstitutionalized subjects aged ≥ 65 years without major neurocognitive impairment at inclusion. At each visit, an orthostatic systolic BP (SBP) ratio was calculated using the formula (sitting SBP/standing SBP) x 100. An orthostatic diastolic BP ratio was computed using the same method. Subjects with a SBP ratio &gt; 100% had sitting SBP &gt; standing SBP. We assessed the visit-to-visit BP postural changes variability according to several indicators: standard deviation, coefficient of variation (CV), variation independent of mean, residual standard deviation, average real variability, and successive variation. Cognitive performances were evaluated using the Mini-Mental State Examination. Linear mixed models were used for the analyses. Results: We included 2,974 subjects (mean age 78 years, women 56%). After adjustment for age, sex, education, baseline body mass index, smoking, alcohol intake, atrial fibrillation, congestive heart failure, transient ischemic attack or stroke, coronary artery disease, type 2 diabetes, dyslipidemia, seated SBP/DBP and antihypertensive treatment, greater visit-to-visit systolic BP postural changes variability was significantly associated with CD over time (adjusted β [95% CI] = -0.17 [-0.29; -0.06], p= 0.004 for CV). Similar results were found for visit-to-visit diastolic BP postural changes variability (adjusted β [95% CI] = -0.14 [-0.25; -0.03], p= 0.017 for CV). Conclusion: Greater visit-to-visit BP postural changes variability was associated with CD. Further studies are needed to assess whether controlling orthostatic BP instability over time could be a promising interventional target to preserve cognition among older adults.

Machine-Learning-Based Surface Ground-Motion Prediction Models for South Korea with Low-to-Moderate Seismicity

ABSTRACT Ground-motion prediction models (GMPMs) have been developed to estimate seismic intensity considering earthquake magnitude, source-to-site distance, site condition, and so on. This study proposes GMPMs to predict 5% damped pseudospectral acceleration (PSA) for 27 periods ranging from 0.01 to 10 s in Korea, based on three machine-learning techniques (i.e., artificial neural network [ANN], random forest [RF], and gradient boosting [GB]). We use 1189 ground motions recorded at 50 surface stations during the 77 earthquakes with a local magnitude (ML) greater than 3.0, including the Gyeongju and Pohang earthquakes with ML of 5.8 and 5.4, respectively. We compare the performances of the three machine-learning-based models and the classical regression-based model in terms of the coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), standard deviation of residuals, and between-event and within-event residuals. The GB-based model shows the best performance. In addition, we analyze the working process of the GB-based model using variable importance and partial dependence (PD) plots. Among the five independent variables (ML, epicentral distance [Repi], average shear-wave velocity of the upper 30 m [VS30], focal depth, and slope angle) used in this study, ML and Repi are the most influential variables and show strong correlations with PSAs. We apply the GB-based model to three recent earthquakes larger than ML 3.0, and the model accurately predicts the PSAs at various stations. We also generate maps of estimated PSA (PSAest) values for the four periods (T = 0.01, 0.1, 1, and 3 s) for the scenario earthquake with an ML of 5.0. We provide a method for training the GB-based model using the Python library, which can enhance the ground-motion prediction not only in Korea but also worldwide, and an executable version of the validated GB-based model.

Exploring an algorithm to harmonize International Obesity Task Force and World Health Organization child overweight and obesity prevalence rates.

SummaryBackgroundThe International Obesity Task Force (IOTF) and World Health Organization (WHO) body mass index (BMI) cut‐offs are widely used to assess child overweight, obesity and thinness prevalence, but the two references applied to the same children lead to different prevalence rates.ObjectivesTo develop an algorithm to harmonize prevalence rates based on the IOTF and WHO cut‐offs, to make them comparable.MethodsThe cut‐offs are defined as age‐sex‐specific BMI z‐scores, for example, WHO +1 SD for overweight. To convert an age‐sex‐specific prevalence rate based on reference cut‐off A to the corresponding prevalence based on reference cut‐off B, first back‐transform the z‐score cut‐offs zA and zB to age‐sex‐specific BMI cut‐offs, then transform the BMIs to z‐scores zB,A and zA,B using the opposite reference. These z‐scores together define the distance between the two cut‐offs as the z‐score difference dzA,B=12zB−zA+zA,B−zB,A. Prevalence in the target group based on cut‐off A is then transformed to a z‐score, adjusted up or down according to dzA,B and back‐transformed, and this predicts prevalence based on cut‐off B. The algorithm's performance was tested on 74 groups of children from 14 European countries.ResultsThe algorithm performed well. The standard deviation (SD) of the difference between pairs of prevalence rates was 6.6% (n = 604), while the residual SD, the difference between observed and predicted prevalence, was 2.3%, meaning that the algorithm explained 88% of the baseline variance.ConclusionsThe algorithm goes some way to addressing the problem of harmonizing overweight and obesity prevalence rates for children aged 2–18.

Evaluation of the Duration of Preanesthesia Consultation: Prospective and Multicenter Study.

The time allocated to the preanesthesia consultation (PAC) of a patient undergoing an elective surgical procedure is an important factor to optimize consultation sessions. The main objective of this study was to build a model predictive of the duration of the PAC. We prospectively studied 1007 patients undergoing a PAC from January 2016 to June 2018 in 4 different hospitals. A general linear model was fitted to predict the overall duration of the PAC. Secondary models predicted the time spent on clinical evaluation and the time assigned to delivering information. After exclusion of 40 patients with major data inconsistencies, the mean (standard deviation [SD]) overall duration of the PAC was 11.2 (5.8) minutes, split into 6.8 (4.1) minutes of information and 4.4 (2.7) minutes of clinical evaluation. It was, respectively, 11.4 (5.9), 6.9 (4.2), and 4.4 (2.7) in the 924 patients ≥16 years of age and, respectively, 8.3 (2.3), 4.3 (1.8), and 4.1 (1.8) in 43 children. The American Society of Anesthesiologists (ASA) score, the number of comorbidities or treatment, surgery discipline, and context (ambulatory, conventional hospitalization, and intensive care unit) were significantly correlated to PAC time. In the 924 adult patients, the models had an R2 adjusted for overfitting at 0.47 for the total duration of PAC, 0.45 for the clinical examination time, and 0.24 for the information time. The estimated residual standard deviations were, respectively, 4.3, 3.1, and 2.7 minutes. The predictive performances of the model explaining the overall duration of PAC were average (R2 = 0.47) and should be confirmed by further studies to use it for optimizing the organization of the consultation by individualizing the time dedicated to each consultation.

Clinical associations of SARS-CoV-2 viral load using the first WHO International Standard for SARS-CoV-2 RNA

SARS-CoV-2 viral load declines from the time of symptom onset; in some studies viral load is higher or persists longer in more severe COVID-19 infection, and viral load correlates with culture positivity. This was a retrospective cohort study of inpatients and outpatients during the first wave of COVID-19 infection in Western Australia, March to May 2020, of the relationship of SARS-CoV-2 viral load (using the First WHO International Standard for SARS-CoV-2 RNA) from symptom onset, by clinical subgroups determined from the public health database and hospital records, using regression analysis. We studied 320 samples from 201 COVID-19 cases: 181 mild, seven severe, 11 critical, and four cases who died (two were also critical cases). At symptom onset the mean viral load was 4.34 log10 IU/mL (3.92–4.77 log10 IU/mL 95% CI, cobas SARS-CoV-2 assay ORF1a Ct 28.9 cycles). The mean viral load change was –0.09 log10 IU/mL/day (–0.12 to –0.06 95% CI). R2 was 0.08 and residual standard deviation 2.68 log10 IU/mL. Viral load at symptom onset was higher for those reporting fever compared to those not reporting fever. Viral load kinetics were not different for gender, age, shortness of breath, or those requiring oxygen. Mean viral load at usual release from isolation at 14 days was 2.5 log10 IU/mL or day 20 was 1.8 log10 IU/mL. Variability in respiratory sample SARS-CoV-2 viral load kinetics suggests viral loads will only have a role supporting clinical decision making, and an uncertain role for prognostication.

Validation of a self-report home food availability checklist against in-home food inventories conducted in low-income Black/African American and Hispanic/Latinx households with preschool-age children

Home food availability (HFA) checklists can be completed by self-report to assess the home food environment. Checklists developed for Black/African American (B/AA) and Hispanic/Latinx (H/L) households are seldom validated against objective approaches such as exhaustive in-home food inventories. This study validated a self-report HFA checklist developed for B/AA and H/L households (n = 97) against researcher-completed HFA checklists verified by exhaustive in-home food inventories. Mean estimates of sensitivity, specificity, and area of the receiver operating curve (ROC), and interrater agreement (Gwet AC1) were calculated to examine the accuracy and agreement of self-reported checklists against direct observation of individual food items. Mean differences in HFA food group scores were compared (self-report vs observed) to examine group-level relative validity. The predictive validity of this self-reported measure on observed scores and dietary intake were also examined with linear regression. The average values for ROC area (average of sensitivity and specificity) ranged from acceptable (0.76 for sweets) to excellent (0.81 for vegetables, fruits). Average interrater agreement values ranged from moderate (0.41–0.60: sweets) to substantial (0.61–0.79: vegetables, fruit, SSBs, savory foods). Self-reported mean scores, compared to observed scores, were higher for vegetables (mean diff: 1.04) and lower for sweets (mean diff: 0.38, p = 0.01), but regression analyses demonstrated that self-reported scores were good predictors of observed scores with absolute error (based on standard deviation of residuals) ranging from ±1.27 to 1.69 points. Self-reported scores also predicted multiple aspects of dietary intake but more so among H/L households. In conclusion, the HFA checklist obtained via self-report performed well based on multiple indicators of validity suggesting that this self-reported measure can be used to assess home food environments among of B/AA and H/L households.

Predicting height growth in bean plants using non-linear and polynomial models

Brazil has stood out worldwide as one of the main producers and consumers of beans, which makes their cultivation important for the economic and social development of the country. As the bean plant has a short growth cycle, its modeling is essential for optimizing management plans for this crop. This modeling can be performed by linear and non-linear models, but the latter have stood out for providing more information to the researcher, mainly due to the practical interpretation of their parameters. In this sense, in the R statistical software, the third-degree linear polynomial model and the Logistic and Gompertz non-linear models were adjusted to height data, in centimeters, in relation to time, in days after emergence, totaling 11 observations. As criteria to assess the quality of the fit, the adjusted coefficient of determination, the corrected Akaike information criterion and the residual standard deviation were used. The logistic model best fitted the data.

Abstract TP222: The Impact Of Il-6 Level On Visit-to-visit Blood Pressure Variability And Incidence Of Stroke: A Post-hoc Analysis Of Mesa

Background: Increased blood pressure variability (BPV) has been associated with stroke risk, but no serum inflammatory markers have been identified associated with increased BPV, despite preclinical data suggesting a role. Methods: This is a post-hoc analysis of the Multi-Ethnic Study of Atherosclerosis (MESA) study. The study exposure was tertiles of serum level of interleukin-6 (IL-6) at the baseline study visit. The primary outcome was visit-to-visit BPV measured as the residual standard deviation (rSD) of at least 4 subsequent study visits (1999-2018). Two logistic regression models were fit to the top tertile (i.e. most variability) of rSD during follow-up: in Model 1, we adjusted for covariates chosen with restricted Least Absolute Shrinkage and Selection Operator (LASSO) and in Model 2, for all covariates. Results: Our analysis included 5,483 patients, of mean (SD) age 61.4 (10.0) years and with 52.9% female. Mean (SD) of IL-6 was 1.7 (1.2) (pg/mL) in the top tertile BPV rSD and 1.4 (1.1) in the bottom tertiles. The incidence of ischemic stroke and all major adverse cardiovascular event was higher in top tertile of BPV (Table 1). Elevated serum IL-6, but not other serum biomarkers of inflammation and clotting (see comment below), was associated with the highest tertile of BPV, independent of potential confounders (Table 2). Conclusion: Higher levels of serum IL-6 were associated with increased subsequent BPV in a large multiracial cohort. Further investigation is needed to better understand the relationship between chronic inflammation and BPV. Table 1. Baseline demographics, serum biomarkers, and clinical outcomes during follow-up (how long was follow-up on average?) in the full cohort, and stratified by the top versus lower tertiles of BPV. Table 2. Association of serum biomarkers at baseline with the highest tertile of blood pressure variability.*Model 1 adjusted for patient age categories (&lt;55, 55-64, 65-69, ≥70), sex, and hypertension. **Model 2 (n=5,447) adjusted for patient age categories (&lt;55, 55-64, 65-69, ≥70), sex, race/ethnicity, education, hypertension, diabetes, smoking, drinking, body mass index, lipid lowering medication, and mean systolic blood pressure

Gaussian process regression based remaining fatigue life prediction for metallic materials under two-step loading

Remaining fatigue life prediction is vital for engineering structures to ensure safety and reliability. It can be more challenging when the structures suffer variable amplitude loadings because of the complex, non-uniform of the fatigue damage accumulation and inherent noise, uncertainty in the data. To further tackle the problem, the Gaussian process regression (GPR) is introduced, which can simultaneously estimate the output value and quantify the associated uncertainty. Therefore, a GPR-based remaining fatigue life prediction method is proposed to predict the remaining fatigue life for metallic materials under two-step loading in this paper. The proposed method is comprehensively evaluated on the dataset containing 12 materials, 328 samples in total. The proposed method achieves the lowest mean square error (MSE), mean absolute percentage error (MAPE), residual standard deviation (RSD) values and the highest correlation coefficient (CC) values among the six machine learning methods and the two model-driven methods. Those results indicate that the proposed method can achieve greater accuracy and reliability in remaining life prediction under two-step loading, which illustrate the effectiveness of the proposed method as a data-driven method in the field of remaining life prediction.

Spatiotemporal functional modeling of postseismic deformations after the 2011 Tohoku-Oki earthquake

Postseismic deformations continue to occur for a long period after major earthquakes. Temporal changes in postseismic deformations can be approximated using simple functions. Since the 2011 Tohoku-Oki earthquake, operating global navigation satellite system stations have been continuously accumulating a remarkable amount of relevant data. To verify the functional model, we performed statistical data processing on postseismic deformations due to this earthquake and obtained their spatiotemporal distribution. Moreover, we approximated the postseismic deformations over a relatively wide area with a standard deviation of residuals of 1 cm for approximately 10 years using a combined functional model of two logarithmic and one exponential functions; however, the residuals from the functional model exhibited a marked deviation since 2015. Although the pattern of postseismic deformations remained unaltered after the earthquake, a change in the linear deformation occurred from 2015 to date. We reduced the overall standard deviation of the residuals of > 200 stations distributed over > 1000 km to < 0.4 cm in the horizontal component by enhancing the functional model to incorporate this linear deformation. Notably, time constants of the functions were similarly applicable for all stations and components. Furthermore, the spatial distribution of the coefficients of each time constant were nonrandom, and the distribution was spatially smooth, with minute changes in the short wavelengths in space. Thus, it is possible to obtain a gridded model in terms of a spatial function. The spatial distributions of short- and long-period components of the functional model and afterslip and viscoelastic relaxation calculated using the physical model were similar to each other, respectively. Each time function revealed a connotation regarding the physical processes, which provided an understanding of the physical phenomena involved in seismogenesis. The functional model can be used to practical applications, such as discerning small variations and modeling for precise positioning.Graphical

Modelos lineares e não lineares para a descrição da curva de lactação da raça Girolando

Abstract The objective of this work was to compare the main linear and nonlinear models used to describe lactation curves and to evaluate the nonlinearity of the nonlinear models, in order to obtain the most adequate model to describe the lactation curves of the Girolando breed. Data from 165 lactations of 89 3/4 Holstein + 1/4 Gyr cows were used, and average yield was calculated every 20 days up to 310 days of lactation. Seventeen models of lactation curves available in the literature were compared. The selection of the best model was based on the curvature measures of Bates &amp; Watts, the bias of Box, adjusted coefficient of determination, Akaike’s information criterion, and residual standard deviation. The linear model of Cobuci estimated a yield peak of 16.7 kg at 40 days of lactation, whereas the nonlinear model of Wood estimated a yield peak of 16.8 kg at 41 days of lactation and a persistence of 6.82. Nonlinearity measures were the most appropriate for selecting the most suitable nonlinear model for the description of lactation curves. To describe the lactation curves of the Girolando breed, the most suitable linear model is that of Cobuci and the nonlinear model is that of Wood.

Descrição da curva de crescimento do fruto do pessegueiro via modelos não lineares sigmoidais difásicos

Abstract The aim of this study was to describe the growth curve of “Aurora 1” peaches using fruit height and diameter data over time through diphasic sigmoidal models constructed from eight combinations of the following models: Brody, Gompertz and Logistic. Data were obtained from an experiment carried out in 2005 in the municipality of Vista Alegre do Alto, São Paulo, Brazil. The parameters of models were adjusted by the least squares method using the Gauss-Newton algorithm implemented in the R software. Assumptions of normality, homogeneity and independence of residues were verified based on Shapiro-Wilk, Breush and Pagan and Durbin-Watson tests, respectively. The goodness of fit of models was verified according to the corrected Akaike information criterion (AICc), residual standard deviation (RSD), asymptote adjustment index (AI) and nonlinearity measures. All models adjusted for both fruit height and diameter variables met the assumptions of normality, independence and homoscedasticity of errors. In addition, all of them present good quality of fit to fruit height and diameter data, since they presented AI values close to one and low RSD values and non-linearity measures. However, the double Gompertz (GG) and the Logistic + Gompertz (LG) models presented, respectively, the best quality of fit to fruit height and diameter data in relation to the other models. It could be concluded that all diphasic sigmoidal models evaluated showed good fit to height and diameter data and can be used to describe the growth curve of “Aurora-1” peaches, according to goodness of fit criteria. However, it is important to highlight that GG and LG models presented the best quality of fit and can be selected to describe the height and diameter growth of “Aurora 1” peach fruits, respectively, with maximum expected growth close to 63 mm in height and 48 mm in diameter.

Growth predictability models in Mangalarga Marchador horses

ABSTRACT: This study evaluated the suitability of the Brody, logistic, and quadratic response plateau models to describe chest and cannon girth data obtained cross-sectionally in Mangalarga Marchador horses, in order to select the best model and predict the growth and maturity of males and females of this breed. Data were collected from 230 horses aged 6 to 176 months, divided by sex and age (16 age classes). The studied models were compared according to each quality evaluator by computing the adjusted coefficient of determination ( R adj 2) and residual standard deviation (RSD) with R statistical software. The chest girths obtained by the models ranged from 172.06 (males) to 181.50 cm (females) (Brody), 172.51 (males) to 181.89 cm (females) (logistic), and 177.67 (males) to 183.09 cm (females) (plateau). For cannon girth, the values were 18.18 (females) to 19.33 cm (males) (Brody), 18.11 (females) to 19.41 cm (males) (logistic), and 18.70 (females) to 19.40 cm (males) (plateau). The logistic model was best suited to describe the growth in chest girth of male and female Mangalarga Marchador horses. For cannon girth growth, the model best suited for males was the logistic model, and the one best suited for females was the Brody model.

Improvement of MiRS Sea Surface Temperature Retrievals Using a Machine Learning Approach

We report on the development of a machine learning approach to improving sea surface temperature (SST) retrievals based on satellite-based microwave channel measurements at frequencies higher than 23 GHz. The approach uses a deep neural network (DNN) trained using Microwave Integrated Retrieval System physical retrievals as inputs and collocated European Centre for Medium-Range Weather Forecasts analyses for training and validation. The DNN was designed to characterize SST retrieval residual and then used to correct the original retrieval. Evaluation based on one year of independent data showed reduction in retrieval residual standard deviation from 3.22 to 1.80 K in January and 3.02 to 1.92 K in July and reduction in mean residual from 0.30 to 0.08 K in January and 0.61 to 0.22 K in July. Comparisons with multilinear regression and machine learning approaches that used measured brightness temperatures as inputs were significantly less effective in retrieving SST directly, although the DNN used brightness temperature also showed improvements. This indicates that physical retrieval provides valuable information useful in characterizing retrieval residual beyond that of the measured radiances. The DNN approach also effectively removed scan angle dependence of retrieval residuals—an important consideration with cross-track instruments. Sensitivity tests indicated that skill declines with time as time increases from training month, but that skill in the same month, one year later is nearly the same as that of the original training month. This suggests that it may be sufficient to pretrain a stratified model with monthly or seasonal dependence using one full annual cycle, which could then be used in subsequent years with continued good performance.

Correction to 'Bayesian inference reveals positive but subtle effects of experimental fishery closures on marine predator demographics'.

Correction to 'Bayesian inference reveals positive but subtle effects of experimental fishery closures on marine predator demographics'.

Shim optimization with region of interest-specific Tikhonov regularization: Application to second-order slice-wise shimming of the brain.

Slice-wise shimming can improve field homogeneity, but suffers from large noise propagation in the shim calculation. Here, we propose a robust shim current optimization for higher-order dynamic shim updating, based on Tikhonov regularization with a variable regularization parameter, . THEORY AND METHODS: was selected for each slice separately in a fully automatic procedure based on a combination of boundary constraints and an L-curve search algorithm. Shimming performance was evaluated for second order slice-wise shimming of the brain at 7T, by simulation on a database of field maps from 143 subjects, and by direct measurements in 8 subjects. Simulations yielded on average 36% reduction in the shim current norm for just 0.4 Hz increase in residual field SD as compared to unconstrained unregularized optimization. In vivo results yielded on average 34.0 Hz residual field SD as compared to 34.3 Hz with a constrained unregularized optimization, while simultaneously reducing the shim current norm to 2.8 A from 3.9 A. The proposed regularization also reduced the average step in the shim current between slices. Slice-wise variable Tikhonov regularization yielded reduced current norm and current steps to a negligible cost in field inhomogeneity. The method holds promise to increase the robustness, and thereby the utility, of higher-order shim updating.