Contribution Of Explanatory Variables Research Articles

Predicting postoperative visual acuity in epiretinal membrane patients and visualization of the contribution of explanatory variables in a machine learning model.

The purpose of this study was to develop a model that can predict the postoperative visual acuity in eyes that had undergone vitrectomy for an epiretinal membrane (ERM). The Light Gradient Boosting Machine (LightGBM) was used to evaluate the accuracy of the prediction and the contribution of the explanatory variables. Two models were designed to predict the postoperative visual acuity in 67 ERM patients. Model 1 used the age, sex, affected eye, axial length, preoperative visual acuity, Govetto's classification stage, and OCT-derived vector information as features to predict the visual acuity at 1, 3, and 6 months postoperatively. Model 2 incorporated the early postoperative visual acuity as an additional variable to predict the visual acuity at 3, and 6 months postoperatively. LightGBM with 100 iterations of 5-fold cross-validation was used to tune the hyperparameters and train the model. This involved addressing multicollinearity and selecting the explanatory variables. The generalized performance of these models was evaluated using the root mean squared error (RMSE) in a 5-fold cross-validation, and the contributions of the explanatory variables were visualized using the average Shapley Additive exPlanations (SHAP) values. The RMSEs for the predicted visual acuity of Model 1 were 0.14 ± 0.02 logMAR units at 1 month, 0.12 ± 0.03 logMAR units at 3 months, and 0.13 ± 0.04 logMAR units at 6 months. High SHAP values were observed for the preoperative visual acuity and the ectopic inner foveal layer (EIFL) area with significant and positive correlations across all models. Model 2 that incorporated the postoperative visual acuity was used to predict the visual acuity at 3 and 6 months, and it had superior accuracy with RMSEs of 0.10 ± 0.02 logMAR units at 3 months and 0.10 ± 0.04 logMAR units at 6 months. High SHAP values were observed for the postoperative visual acuity in Model 2. Predicting the postoperative visual acuity in ERM patients is possible using the preoperative clinical data and OCT images with LightGBM. The contribution of the explanatory variables can be visualized using the SHAP values, and the accuracy of the prediction models improved when the postoperative visual acuity is included as an explanatory variable. Our data-driven machine learning models reveal that preoperative visual acuity and the size of the EIFL significantly influence postoperative visual acuity. Early intervention may be crucial for achieving favorable visual outcomes in eyes with an ERM.

What explains gender inequality in HIV infection among high-risk people? A Blinder-Oaxaca decomposition

BackgroundDespite clear evidence on role of gender in vulnerability and exposure to HIV infection, information on gender-related inequalities in HIV and related factors are rarely documented. The aim of this study was to measure gender inequality in HIV infection and its determinates in Tehran city, the capital of Iran.MethodsThe study used the data of 20,156 medical records of high-risk people who were admitted to Imam Khomeini Voluntary Counseling and Testing site in Tehran from 2004 to 2018. The Blinder-Oaxaca decomposition was used to quantify the contribution of explanatory variables to the gap in the prevalence of HIV infection between female and male.ResultsThe age-adjusted proportion of HIV infection was 9.45% (95%Cl: 9.02, 9.87). The absolute gap in the prevalence of HIV infection between male and female was 4.50% (95% CI: − 5.33, − 3.70%). The Blinder-Oaxaca decomposition indicated that most explanatory factors affecting the differences in HIV infection were job exposure, drug abuse, history of imprisonment, injection drug, heterosexual unsafe sex, and having an HIV-positive spouse.ConclusionThe results can provide evidence for health policymakers to better planning and conducting gender-based preventive and screening programs. Policies aiming at promoting HIV preventive behaviors among male may reduce the gap in HIV infection between female and male in Iran.

Phylogenetic conservatism explains why plants are more likely to produce fleshy fruits in the tropics.

Plant functional traits often show strong latitudinal trends. To explain these trends, studies have often focused on environmental variables, correlations with other traits that themselves show latitudinal trends, and phylogenetic conservatism. However, few studies have systematically disentangled the relative contributions of these factors. Using a dataset consisting of 9,370 plant species from Southwest China, we investigated factors affecting fruit type (fleshy vs. dry): plant growth form, environmental constraints (summarized by climate region), and phylogenetic conservatism. Growth form and climate region are often cited in the literature as important explanations for the higher proportion of fleshy fruited species in the tropics. Nonetheless, in our analyses using partial R2 , growth form and climate region explained only 1.7% and 0.3%, respectively, of the variance in fruit type in a model including phylogeny, while phylogenetic conservatism explained 79.5%. Furthermore, phylogenetic conservatism was evenly distributed along the phylogeny, implying that fruit type reflects both ancient and recent phylogenetic relationships. Our findings illustrate the value of parsing out the contributions of explanatory variables and phylogeny to the variance in species' traits. Methods using phylogenies that calculate partial R2 give a more informative tool than traditional methods to explore the phylogenetic patterns of functional traits.

Disentangling the Contributions of Climate and Basin Characteristics to Water Yield Across Spatial and Temporal Scales in the Yangtze River Basin: A Combined Hydrological Model and Boosted Regression Approach

The dependence and contribution of explanatory variables or predictors to water yield need to be closely analyzed and accurately quantified to better understand water balances as well as for effective water resources management. It is generally challenging, however, to disentangle the contribution of individual climate variables from that of basin characteristics to the integrated water yield response. Here we propose a method to concurrently quantify and analyze the effects of climate and basin predictors on water yields. This method employs the Soil and Water Assessment Tool (SWAT) to simulate water yield. Simulated results are then analyzed and compared using Boosted Regression Trees (BRTs) at multiple spatial and temporal scales. Results indicate that in the Yangtze River Basin (YRB) on average, precipitation is of paramount importance, followed by land cover, while slope has the lowest contribution. The average relative contributions of soil moisture, maximum and minimum temperatures are different among temporal scales. More stable and reliable results are derived at the daily scale compared to the yearly and monthly scale. Our results make evident that generalizations about water yield response made in the absence of a comprehensive and accurate description of site- and scale-specific contributions can lead to misleading assessments. This proposed approach can be useful for informing and supporting more effective water resources management goals.

Public funding of pharmaceuticals in the Netherlands: investigating the effect of evidence, process and context on CVZ decision-making

The College Voor Zorgverzekeringen (CVZ) provides guidance to the Dutch healthcare system on funding and use of new pharmaceutical technologies. This study examined the impact of evidence, process and context factors on CVZ decisions in 2004-2009. A data set of CVZ decisions pertaining to pharmaceutical technologies was created, including 29 variables extracted from published information. A three-category outcome variable was used, defined as the decision to 'recommend', 'restrict' or 'not recommend' a technology. Technologies included in list 1A/1B or on the expensive drug list were considered recommended; those included in list 2 or for which patient co-payment is required were considered restricted; technologies not included on any reimbursement list were classified as 'not recommended'. Using multinomial logistic regression, the relative contribution of explanatory variables on CVZ decisions was assessed. In all, 244 technology appraisals (256 technologies) were analysed, with 51%, of technologies recommended, 33% restricted and 16% not recommended by CVZ for funding. The multinomial model showed significant associations (p ≤ 0.10) between CVZ outcome and several variables, including: (1) use of an active comparator and demonstration of statistical superiority of the primary endpoint in clinical trials, (2) pharmaceutical budget impact associated with introduction of the technology, (3) therapeutic indication and (4) prevalence of the target population. Results confirm the value of a comprehensive and multivariate approach to understanding CVZ decision-making.

The Role of Risk and Transaction Costs in Contract Design: Evidence from Farmland Lease Contracts in U.S. Agriculture

AbstractThe objective of this article is to provide new empirical evidence on landlord‐tenant choices of share versus cash‐rent contracts in U.S. agriculture. The focus is on the contribution of explanatory variables that represent transaction costs, risk‐sharing incentives, or both. An empirical model of contract choice is tested against the 1999 Agricultural Economics and Land Ownership Survey (AELOS) and finds mixed evidence for low transaction cost and risk‐sharing‐incentive motives for landlord‐tenant choices of a share versus cash‐rent contract. However, the behavior of landlords and tenants is consistent with them being risk averse. Although it is standard to control for the riskiness of the principal's task that is contracted, we find that other attributes of the landlord are an important part of a relatively complex story for U.S. land tenancy contacting. The latter results have generally been ignored in other published landlord‐tenant contracting studies.

Determinant variables of Iberian Peninsula Aphodiinae diversity (Coleoptera, Scarabaeoidea, Aphodiidae)

AbstractAim The aims of this paper are to examine diversity–variability patterns for species of Aphodiinae (Coleoptera, Scarabaeoidea, Aphodiidae) on the Iberian Peninsula, and to determine the factors that influence their geographic distribution.Location Iberian Peninsula (Spain and Portugal).Methods Data from 30 studies and their bibliographies on species of Iberian Peninsula Aphodiinae were compiled. The reliability of the inventories was evaluated using parametric species richness estimators. In addition, a further 11 variables related to rarity, geographic distribution, or phylogenetic diversity were considered. Diversity variables were analysed using principal components analysis to reduce the number of dependent variables. Subsequently, the effect of differences in locality size among the 30 studies was eliminated by calculating and retaining the residuals of the curvilinear relationship of each diversity variable with the area. Generalized linear models were used to examine the relationships between diversity and 17 environmental variables. The diversity variables and their residuals were also subject to trend surface analysis in order to identify the relevance of spatially structured variables that had not been considered. The contribution of explanatory variables was determined through hierarchical variance analysis.Results Principal components analysis of biodiversity variables revealed that most of the variability could be explained using three biodiversity indexes: BI1, correlated positively with species richness, widely distributed species, frequent species, abundant species, species occurring in North Africa, Europe and the Iberian Peninsula, and phylogenetic diversity; BI2, correlated positively with numbers of infrequent and African–Iberian species; and, BI3, correlated positively with numbers of endemic, non‐abundant, European, and Iberian‐restricted species. A latitudinal disjunction emerged in BI1, with maximum scores at the north‐western and southern corners, while maximum BI2 scores were found throughout the south, and maximum BI3 scores in the north‐west. For BI1, it was climate that had the greatest influence, followed by lithology, and livestock presence. Geographic variables were the most significant for BI2, followed by climate and livestock presence. Finally, for BI3, climate variables were the most important, while geography, lithology and livestock presence had some relevance.Main conclusions The relevance of geographic variables indicates that other unaccounted‐for factors that are spatially structured could possibly explain additional variation in Aphodiinae diversity. These factors may be historic in nature, relating to the species groups, namely the Ibero‐European and the Mediterranean or Afro‐Iberian. The northern pattern could reflect the fact that the Iberian Peninsula acted as a colonization route and as a refuge during the glacial/interglacial cycles, while the southern pattern could be a consequence of the connection between the Iberian Peninsula and North Africa during the Messinian crisis, and/or a historic relationship in common, related to human activity.

Multiple Correlation by Means of the Linked Vector Pattern Based on Ranked Data

In this paper, we propose a linked vector multiple correlation analytical graph, where given data are transformed to ranked ones, and propose a new measure which expresses the degree of multiple correlation quantitatively. Weights of contribution of explanatory variables to an objective variable can be determined from this measure. This measure is considered to be reasonable by comparing both with the results of the method by human eyes, and with the one of multiple regression analysis for data which are transformed to ranked data.