Important Explanatory Variables Research Articles

Residential building and sub-building level flood damage analysis using simple and complex models

AbstractFlood damage assessment is critical for optimal risk management investments. Damage models evaluate physical damage or monetary loss from direct building exposure to flood hazard processes. Traditional models represent a simple relationship whereby physical damage increases with water depth. More complex models offer an improved understanding of vulnerability, analysing interactions between multiple hazard and exposure variables that drive damage. Our study investigates whether increasing model complexity and explanatory damage variables improves prediction precision and reliability at residential building and sub-building (component) levels. We evaluate simple and complex empirical univariable and multivariable models for flood damage prediction. The Random Forest algorithm learned on multiple hazard and exposure explanatory variables outperformed linear and non-linear univariable regression approaches. Random Forest model predictive precision was highest when learning was limited to water depth and several important explanatory damage variables (flow velocity, area and floor height). Component damage models demonstrated high predictive precision for internal finishes and services. Precision reduced for structure and external finishes as damage samples for model learning were limited. High performing but complex multivariable models require further spatio-temporal transfer investigation to determine opportunities for accurate and reliable object-specific flood damage prediction in data scarce locations.

Psychological Characteristics and Quality of Life of Patients With Functional Dyspepsia.

The objective of this study is to compare the psychosocial characteristics of functional dyspepsia (FD) with its subgroups, epigastric pain syndrome (EPS) and postprandial distress syndrome (PDS), against a healthy control group, and to investigate the quality of life (QoL). All of the subjects were 210 adults, 131 patients with FD were diagnosed by gastroenterologist and 79 adults with no observable symptoms of FD were selected as the normal control group. Demographic factors were investigated. The Korean-Beck Depression Inventory-II, Korean-Beck Anxiety Inventory, Korean-Childhood Trauma Questionnaire, Multidimensional Scale of Perceived Social Support, Connor-Davidson Resilience Scale, and WHO Quality of Life Assessment Instrument Brief Form were used to assess psychological factors. A one-way analysis of variance was used to compare differences among the groups. Further, a stepwise regression analysis was conducted to determine factors affecting the QoL of the FD group. Between-group differences in demographic characteristics were not significant. Depression (F=37.166, p<0.001), anxiety (F=30.261, p<0.001), and childhood trauma (F=6.591, p<0.01) were all significantly higher in FD group compared to the normal control. Among FD subgroups, EPS exhibited higher levels of both depression and anxiety than PDS. Social support (F=17.673, p<0.001) and resilience (F=8.425, p<0.001) were significantly lower in FD group than in other groups, and the values were higher in PDS than in EPS. Resilience (β=0.328, p<0.001) was the most important explanatory variable. The explained variance was 46.6%. Significantly more symptoms of depression, anxiety, childhood trauma was observed for both FD sub-group. These groups also had less social support, resilience, and QoL than the control groups.

Review and meta-analysis of correlates of home range size in bats

Abstract Home range is an ecological concept that affects many aspects of the life of vertebrates and hence understanding how it varies between species is crucial. Mammalian home range size has been linked to body size and diet, but these studies were based predominantly on terrestrial species and most specifically excluded bats. As the only group of flying mammals, bats experience distinctly different constraints on movement, and hence home range. However, despite their diversity, relatively little is known about the home ranges of bats, and this is the first global review assessing correlates of their home range size. Our hypothesis is that home range will be impacted by different ecological conditions experienced by the bats and by the biological traits of the bats themselves. We performed a meta-analysis based on published data for 81 bat species to identify variables that contribute most to home range size. Sex, wing loading, functional group, colony size, dietary class, distance from the equator (latitudinal region), habitat type, and the interaction between habitat type and latitudinal region were all important explanatory variables. Wing loading was positively correlated with home range size—while females, open-air foragers, large colony sizes, and bats in temperate regions consistently had large home ranges. Understanding the correlates of home range has important implications, for example, for bat conservation and for assessing the risk of spillover of zoonotic pathogens from bats to humans and livestock.

Empirical Evidence on the Stock–Bond Correlation

The correlation between stock and bond returns is a cornerstone of asset allocation decisions. History reveals abrupt regime shifts in correlation after long periods of relative stability. We investigate the drivers of the correlation between stocks and bonds and find that inflation, real rates, and government creditworthiness are important explanatory variables. We examine the implications of a shift in the stock–bond correlation and find that increases are associated with higher multi-asset portfolio risk and higher bond risk premia.

Abstract PO1-02-14: Breast cancer in older women - long-term prognosis by age and subtype in a large population-based cohort

Abstract Background Compared to middle-aged women, older women have a higher risk of breast cancer death. Whether the increased risk of breast cancer death is present in all subgroups of older adults with breast cancer, or negligible in the subgroups receiving adequate surgical and oncological treatment, has been debated. We studied long-term prognosis by age and subtype, adjusting for stage, treatment, socioeconomic status and comorbidity, clarifying which subgroups of older patients that are at increased risk of breast cancer death. Methods In a nation-wide cohort of 45,949 women aged ≥60 years at diagnosis of breast cancer 1997-2014, we estimated crude probabilities of breast cancer death and from other causes using competing risk methods. Cox proportional hazards models were used to estimate adjusted hazard ratios (HR) with 95% confidence intervals (CI. Results In this large population-based cohort study of older women with breast cancer, we found an increasing risk of breast cancer death with increasing age at diagnosis, both at 5- and 10-year follow up. The strong association between age and breast cancer death was found for all breast cancer subtypes, being most prominent in Her2-positive breast cancer. The age effect was also significant in Luminal A breast cancer where treatment opportunities are not necessarily affected by age. High-risk tumors were equally common in all age groups above 60 years. In multivariable analysis, stage, subtype, grade and treatment were important explanatory variables to the age-related differences in survival, while comorbidity and socioeconomic status contributed minorly. After adjustments, older age remained an independent risk factor for both overall and breast cancer survival. Older women with triple-negative (60-79 years) or Her2-positive (60-84 years) tumors had a 5-year risk of breast cancer death exceeding the risk of death by other causes. Conclusions Older women had a significantly higher risk of breast cancer death irrespective of breast cancer subtype, also in Luminal A disease where treatment opportunities are not necessarily affected by age. Table 3. Crude probabilities of breast cancer specific death and death by other cause (competing cause of death) at 5, 10 and 15 years. Table 4. Multivariable analysis of prognostic factors affecting 15-year overall (OS) and breast cancer- specific survival (BCSS) Table 5. Multivariable analysis of prognostic factors affecting 15-year overall (OS) and breast cancer- specific survival (BCSS) by age at diagnosis and subtype. Citation Format: Amelia Chiorescu, Anna L V Johansson, Antonis Valachis, Alexios Matikas, Jan Frisell, Lars Holmberg, Irma Fredriksson. Breast cancer in older women - long-term prognosis by age and subtype in a large population-based cohort [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO1-02-14.

Quantitative evaluation and mechanism analysis of soil chemical factors affecting rice yield in saline-sodic paddy fields

Salinization and sodication have become an important abiotic stress affecting soil fertility and crop production in the western of the Songnen Plain in Northeast China. And rice cultivation is considered as one of the most effective biological methods to reclaim saline-sodic soils and ensure food security. However, it is difficult to select the optimal measures to regulate rice growth for increasing yield, because the independent and comprehensive influences of the soil limitation factors on rice yield are not quantitatively evaluated. In this study, the hierarchical partitioning (HP) and the structural equation model (SEM) were used to quantitatively evaluate the influences of salinization parameters, salt ion concentrations and soil nutrients to identify the dominant limitation factors and obstacle mechanism for rice yield. The results showed that soil pH was the key index in salinization parameters, [CO32− + HCO3−] was the key index in salt ion concentrations and available nitrogen (AN) was the key index in soil nutrients to impact rice yield, which independent influences reached 53.7 %, 45.4 % (negative) and 53.2 % (positive), respectively. Soil pH was determined by [CO32− + HCO3−], and the negative effect of alkali stress on rice yield mainly caused by [CO32− + HCO3−] was greater than that of salt stress mainly caused by [Na+] in saline-sodic paddy fields. Among the soil chemical factors, soil pH and AN were the most important explanatory variables of rice yield in saline-sodic paddy fields, which standardized total effects were − 0.32 and 0.40, respectively. Furthermore, the AN showed a more significant negative correlation with soil pH and a higher yield-increasing potential in severe saline-sodic soils (9 ≤ pH < 10) than that in moderate saline-sodic soils (8 ≤ pH < 9). Therefore, decreasing [CO32− + HCO3−] and increasing the content of AN are key to improve rice yield in saline-sodic paddy fields.

Forecasting East and West Coast Gasoline Prices with Tree-Based Machine Learning Algorithms

This study aims to forecast New York and Los Angeles gasoline spot prices on a daily frequency. The dataset includes gasoline prices and a big set of 128 other relevant variables spanning the period from 17 February 2004 to 26 March 2022. These variables were fed to three tree-based machine learning algorithms: decision trees, random forest, and XGBoost. Furthermore, a variable importance measure (VIM) technique was applied to identify and rank the most important explanatory variables. The optimal model, a trained random forest, achieves a mean absolute percent error (MAPE) in the out-of-sample of 3.23% for the New York and 3.78% for the Los Angeles gasoline spot prices. The first lag, AR (1), of gasoline is the most important variable in both markets; the top five variables are all energy-related. This paper can strengthen the understanding of price determinants and has the potential to inform strategic decisions and policy directions within the energy sector, making it a valuable asset for both industry practitioners and policymakers.

Response of soil moisture to prewinter conditions and rainfall events at different distances to gully banks in the Mollisol region of China

AbstractSoil moisture patterns are crucial factors in gully bank erosion and may significantly vary depending on the water conditions before winter freezing. Rainfall events can evoke different moisture responses throughout the soil profile of gullies. However, the relative importance of prewinter conditions and rainfall events to the soil moisture responses of gully banks in seasonally frozen regions remains unclear. In this study, a field infiltration experiment of a gully bank with four prewinter soil moisture regimes and three replicates was conducted in October 2017 in the Mollisol region of China. We aimed to elucidate the temporal variations in moisture responses under different prewinter water treatments at various distances to the gully bank after three representative rainfall events in 2018. The response magnitude (∆s) varied widely and was largest in the intermediate layers (30–50 cm) for wetter prewinter regimes (i.e., HI and MED) but was larger in the upper soil layers (0–30 cm) for drier prewinter regimes (i.e., CK and LO). In contrast to the findings of other studies, we found that the relative influences of prewinter conditions and rainfall events on the soil moisture responses varied with soil depth and distance to the gully edge. The soil moisture response increased with rainfall depth, and the effect of prewinter conditions dominated until the end of the rainy season (August). The most important explanatory variables for the response magnitude (∆s) were rainfall events (46%), soil depth (10%), and prewinter conditions (7%). This study increases the understanding of the heterogeneity in soil moisture responses and the relationship to rainfall events under different prewinter conditions on gully banks in the Mollisol region. Understanding subsurface hydrological processes and identifying the relative importance of prewinter conditions and rainfall are beneficial for building gully bank erosion models for seasonally frozen regions.

Space use of invertebrates in terrestrial habitats: Phylogenetic, functional and environmental drivers of interspecific variations

AbstractAimWe present the first global database of movement patterns of terrestrial invertebrates, focusing on active dispersal and foraging movements. We depict interspecific variations in movement distances among invertebrates and assess potential drivers of these variations. We finally contrast our results with those of previous vertebrate studies.LocationWorldwide.MethodsWe conducted a meta‐analysis using 176 studies. They provided 411 movement estimates: 175 foraging movement estimates for 101 species, 51 families and 19 orders, and 236 dispersal estimates for 131 species, 53 families and 16 orders. These estimates were complemented by several organism functional traits: body mass, diet, locomotion mode and by environmental variables of the study sites: temperature and NDVI. We computed allometric relationships between movement distances and body mass both globally and separately for each taxonomic order with sufficient data. We tested the relative influence of the co‐variables on movement distances through model selection.ResultsWe reveal a general positive allometric relationship between movement distance and body mass that holds across most taxonomic orders. We evidence a strong phylogenetic signal in movement distances that translates in variable allometries of movement distances with body mass across taxonomic orders. We further find that interspecific variations of movement distances are primarily correlated with functional differences rather than environmental conditions. Locomotion mode is the most important explanatory variable of both dispersal and foraging distances, with larger distances among flying individuals followed by walking and crawling ones for a given body mass. Surprisingly, trophic guild has a low predictive power of invertebrate movement distances as opposed to vertebrates, with unclear differences between invertebrate carnivores, herbivores and decomposers.Main ConclusionsOur study provides general allometric equations for terrestrial movement distances of invertebrates. It further reveals important functional drivers of their interspecific variation in space use with a dominant role of their evolutionary history.

개인의 요구와 비영리 조직의 정보 요인이 기부 행위에 미치는 영향요인에 관한 연구

This study focuses on factors influencing individual charitable donations, specifically examining individual demands and organizational information factors. Previous studies has shown interest in various individual factors and nonprofit organizational efforts affecting charitable donations and volunteerism. However, there has been limited scholarly interest in the impact of individual demands and organizational information factors as influential factors for charitable giving. This study explores the influence of individual efforts to acquire organizational information and express individual opinions and the organization's ability to accept individual opinions and provide information on donation behavior. The results suggest that both individual demands and non-profit organizational information factors are important explanatory variables that can predict individual charitable donations. Finally, this study showed that trust in nonprofit organizations is identified as the most powerful explanatory variable for individual charitable giving.

Assessing the measuring time of forest plots

Forests provide a wide range of ecosystem services and information requirements on forests have grown considerably. Nevertheless, collecting information in the forest is expensive and for this reason assessment of forest resources strongly relies on statistical sampling. However, plot measurement remains essential even when remote-sensing data are used, and field assessments are still among the costliest components of forest inventories. Studies on the costs for plots survey are limited and usually based on expert evaluation rather than on data. This article analysed the relationship between the time needed for measuring forest plots and their site-related characteristics (slope, terrain roughness), stand features (number of trees, subsample trees, stumps, coarse woody debris, understorey vegetation) and protocol-related procedures, by means of univariate and multivariate analyses. Analyses showed that the time needed for measuring plots depends on the workload or the intensity of fieldwork. Especially, the number of variables surveyed matters, because the variables explained the measuring time variation by an additive effect, suggesting that within a complex field protocol the total number of measurements taken may not represent properly the overall intensity of work. Marking in an effective way permanent plots is recommendable because the retrieval success of the pins buried into the ground was the most important explanatory variable. Presence of understorey vegetation was more important than the number of individuals measured. The results obtained are consistent and logical, but the variance explained was limited, suggesting that predictability of the measuring time under complex field protocols might be intrinsically limited by the interactions among factors with opposite effects and especially by the adaptation of the surveyors to specific circumstances. For example, effects of physical tiredness were not detected in the days when two plots were measured; conversely, measuring a second plot reduced the measuring time of the first, an event most likely dependent on the surveyors’ behaviour. Under the conclusion that predictability is low by nature, the inference of studies based on simulation

Comparable diameter resulted in larger leaf area and denser foliage in the park trees than in street trees: A study on Norway maples of Karlsruhe city, Germany

The leaf area of trees is the main surface of energy and matter exchange between the plant canopy and the atmosphere. A better understanding of canopy structure variations in cities is a prerequisite to evaluating leaf area and ecosystem services. We selected 58 single-standing and healthy Norway maple trees (Acer platanoides L.), equally distributed between the park and street at the same tree size for canopy structure and leaf area measurements. The canopy structures of street trees and park trees were different. Street trees had significantly higher dieback (p < 0.01), crown damage (p < 0.01), and branch-free bole length (p < 0.001) than park trees. Even though we sampled trees with similar diameters, park tree crowns tended to be healthier and denser than street trees. The crown volume, crown projection area (CPA), light availability (Crown Light Exposure or CLE), and foliage density were lower in street trees than in park trees. The average foliage density of street trees is 20 % lower than park trees. All the above differences in crown volume and foliage density lead to a significantly lower leaf area in street trees. The total leaf area of a single street tree was only 83 m2 on average, compared to 186 m2 among park trees. We demonstrated that crown volume and growing habitats (i.e., park or street) are important explanatory variables for leaf area. We conclude that a precise and site-specific evaluation of leaf areas is a prerequisite for accuracy in quantifying ecosystem services from urban trees.

The effects of presidential breakdowns on foreign policy in Latin America

AbstractPresidents and their ideologies have long been thought to be important explanatory variables in Latin American foreign policy, but little is known about how presidential breakdowns, a common form of leader change, affect foreign policy. I use a qualitative structured comparison to analyze the effects of nine presidential ousters on their successors’ foreign policies. I focus on cases in which the legitimacy of the process is in a gray zone, when narrative disputes are more likely to emerge. The findings indicate that incumbents politicize foreign policy when they use it to attain public support but accommodate it to congressional interests when they need support of Congress; meanwhile, international peers respond to presidential breakdowns according to their strategic and ideological rather than normative concerns. Even though Latin America is dominated by weak states, domestic politics, especially during crises, play a major role in explaining their foreign policies. Considering the role played by presidents in foreign policymaking, understanding this kind of presidential succession is essential.

International trade and economic growth in Croatia

This article argues that Croatia’s growth over the past two decades is deeply related to the dynamics of international trade. Under the premise that what is bought and sold in global markets reflects the economy’s fundamentals, we show that the rate of growth compatible with equilibrium in the balance-of-payments, i.e. the dynamic Harrod trade multiplier, is a good predictor of the country’s actual long-run growth rate. For this purpose, we apply a state–space model and the Kalman smoother to obtain time-varying parameter estimates of the exports and imports functions. We use these estimates to investigate the determinants of international non-price competitiveness. Bayesian Model Averaging (BMA) and Weighted Average Least Squares (WALS) techniques are combined to tackle model selection uncertainty. It is shown that R&D investments and human capital accumulation are the most important explanatory variables. The success of the European Union’s (EU) integration process depends on the capacity of its new members to achieve a balanced-growth path. Being the last country to join the union, the experience of this single country is particularly relevant to understand possible development alternatives for the region. We conclude by highlighting the policy relevance of our findings to the evaluation of Croatia’s catching-up performance as part of the EU.

Occurrence and removal prediction of pharmaceuticals positively correlated with antibiotic resistance genes in wastewater treatment processes

Wastewater treatment plants (WWTPs) are important source of pharmaceuticals in the environment, identifying their relationship with antibiotic resistance genes (ARGs) and removal behavior is important for risk control. In this study, the correlations of common pharmaceuticals (including antibiotics and non-antibiotics) and ARGs both in wastewater and sludge were investigated, and machine learning was applied to predict the removal efficiencies of typical pharmaceuticals positively correlated with ARGs. Among the 23 kinds of target pharmaceuticals, both antibiotics and non-antibiotics were found to be positive with typical ARGs in wastewater (e.g. correlation coefficients between caffeine and blaTEM, aadD and qnrS were 0.738, 0.609, 0.936, p < 0.01) but negative in sludge (e.g. correlation coefficients between ofloxacin and tetO, tetW and qnrS were −0.922, −0.933 and −0.902, p < 0.01), indicating the high risk of pharmaceuticals in wastewater promoting ARG spread. Furthermore, based on pharmaceutical removal efficiencies in different wastewater treatment processes in this and previous studies, the removal efficiencies of typical pharmaceuticals highly correlated with ARGs were well predicted through process operation parameters and wastewater characteristics by Random Forest model. HRT and temperature were identified as the most important explanatory variables. This study could provide comprehensive references for controlling pharmaceuticals in WWTPs.

Does non‐native diversity mirror Earth's biodiversity?

AbstractAimHuman activities have introduced numerous non‐native species (NNS) worldwide. Understanding and predicting large‐scale NNS establishment patterns remain fundamental scientific challenges. Here, we evaluate if NNS composition represents a proportional subset of the total species pool available to invade (i.e. total global biodiversity), or, conversely, certain taxa are disproportionately pre‐disposed to establish in non‐native areas.LocationGlobal.Time periodPresent day.Major taxa studiedGlobal diversity.MethodsWe compiled one of the most comprehensive global databases of NNS (36,822 established species) to determine if NNS diversity is a representative proportional subset of global biodiversity.ResultsOur study revealed that, while NNS diversity mirrors global biodiversity to a certain extent, due to significant deviance from the null model it is not always a representative proportional subset of global biodiversity. The strength of global biodiversity as a predictor depended on the taxonomic scale, with successive lower taxonomic levels less predictive than the one above it. Consequently, on average, 58%, 42% and 28% of variability in NNS numbers were explained by global biodiversity for phylum, class and family respectively. Moreover, global biodiversity was a similarly strong explanatory variable for NNS diversity among regions, but not habitats (i.e. terrestrial, freshwater and marine), where it better predicted NNS diversity for terrestrial than for freshwater and marine habitats. Freshwater and marine habitats were also greatly understudied relative to invasions in the terrestrial habitats. Over‐represented NNS relative to global biodiversity tended to be those intentionally introduced and/or ‘hitchhikers’ associated with deliberate introductions. Finally, randomness is likely an important factor in the establishment success of NNS.Main conclusionsBesides global biodiversity, other important explanatory variables for large‐scale patterns of NNS diversity likely include propagule and colonization pressures, environmental similarity between native and non‐native regions, biased selection of intentionally introduced species and disparate research efforts of habitats and taxa.

The number of syllables per phrase in Great Tit (Parus major) song decreases in strong anthropogenic noise and at northern latitudes

Cover photo: Great Tit male. Photo: Alf Tore Mjøs. Birds may sing to defend a territory and to attract a mate. However, despite many studies clear conclusions remain on how ecological conditions affect the song, such as physical obstacles that may reduce the sound transmission, and anthropogenic noise that may mask the signal. The social environment of the local populations may also be important, such as breeding density and sex ratio, influencing the number of competing males with which to song match, and the distances to the neighbouring males and to prospecting females. During 2016-19, we counted the number of syllables (notes) per phrase of singing male Great Tits Parus major by visiting seven countries in Europe and one in North Africa. A total of 946 songs were observed by visiting 554 territories. We also recorded study year, anthropogenic noise, calendar date, time of day, type of habitat and vegetation density, latitude, longitude and altitude. The most important explanatory variables were anthropogenic noise and the latitude of the focal site; the number of syllables per phrase decreasing both with increasing anthropogenic noise and with the latitude. The latter result was also supported when analysing sonograms of the species found on the Internet (Xeno-canto), namely fewer syllables per phrase in Norway than in Spain and Portugal. We suggest that repetition of a short phrase is fast interpreted by conspecifics in noisy environments, and that such signals are more readily detected by conspecifics over a wider area where the density of the tits is low.

Environmental drivers of yellowtail kingfish, Seriola lalandi, activity inferred through a continental acoustic tracking network

Abstract Identifying the species response to changing environments can contribute towards proactive and adaptable resource management and, although obtaining observations can be logistically challenging for aquatic species, can be postulated through monitoring. A network of acoustic tracking receivers (n = 93) across south‐eastern Australia was used to identify the effects of environmental conditions on the activity of the yellowtail kingfish (Seriola lalandi, n = 63), an economically important species with a crucial role in pelagic ecosystems. Activity (measured via tri‐axial acceleration) provides an insight into the energetic expenditure of animals, which is linked to movement, behaviour, and physiological processes. Kingfish activity was strongly influenced by sea surface temperature and hour of day, with smaller effects from distance to nearest landmass and bathymetry. Activity also decreased during higher tides and periods of greater moon fraction. Findings show that the energetic responses of kingfish are sensitive to long‐ and short‐term changes, which can regulate behaviours and physiological processes. Changes in kingfish activity and movement (residency and space use) were further investigated at a seasonal aggregation in a small temperate estuary (approx. 120 km2; Coffin Bay, South Australia), where individuals remained during the austral spring and summer (September–April), with a complete exodus in winter. Fifty per cent of tagged fish returned to this estuary in three consecutive years, indicating its importance for aggregating kingfish. While residing in Coffin Bay, kingfish activity varied between interconnected areas, with temperature, hour of day, tide height, and moon fraction again identified as important explanatory variables. These findings have implications for the energetic budgets of large pelagic fish in subtropical and temperate regions, which are facing rapidly changing climates. These results are important for understanding and accounting for the potential responses and physiological impacts of future climatic conditions on migratory pelagic species.

Determinants of teachers’ positive perception on their professional development experience: an application of LASSO-based machine learning approach

ABSTRACT Given the complex nature of teachers’ professional development (PD) processes, it is crucial to examine how various factors surrounding teachers are associated with the evaluation of their PD experience. By applying a machine-learning technique, least absolute shrinkage and selection operator (LASSO), we were able to include numerous factors in an integrated model to create a data-driven, parsimonious predictive model that is readily applicable. Using TALIS 2018 U.S. data (n = 2,418), we identified 16 important explanatory variables (out of 132 variables) in determining teachers’ positive perception on their PD. We found that teachers’ PD experience depends on multiple layers of factors such as features of PD activities (10 variables), teachers’ individual characteristics (four variables), and school organisational environments (two variables). Theoretical and practical implications are also discussed.

머신러닝을 활용한 4년제 대학 중도탈락률 예측요인 탐색

The purpose of this article was to identify factors that predict the dropout rate in 4-year universities using machine learning algorithms. The Random Forest and XGBoost algorithms were employed, and data from the Korean Council for University Education and the University Information Disclosure System were utilized for model development. The collected data spanned from 2011 to 2021 and included a total of 17 independent variables related to educational finance, educational conditions, and educational and research performance. The dropout rate was analyzed as the dependent variable. The final Random Forest and XGBoost models demonstrated the ability to explain 65% and 66% of the variability in the dropout rate, respectively. Using the two machine learning models, the top 10 explanatory variables were identified based on their importance, excluding the student enrollment rate. The models identified the following variables as important predictors of the dropout rate: expenditure per student for academic materials, research budget per full-time faculty member within the university, research budget per full-time faculty member outside the university, freshman enrollment rate, employee-to-student ratio, and employment rate. By examining the partial dependence plots of the five common important explanatory variables, it was analyzed how the predicted dropout rate changes in response to variations in these variables. Although there were differences in the curves between Random Forest and XGBoost, the overall direction and trends of the curves for the important explanatory variables were found to be similar. The improvement of these five important variables was found to have a negative correlation with the dropout rate, indicating a decrease in the predicted dropout rate. Based on these research findings, key implications and suggestions for further studies were provided.