Refine
Has Fulltext
- no (21) (remove)
Document Type
- Article (21) (remove)
Language
- English (21)
Is part of the Bibliography
- yes (21)
Keywords
- model (21) (remove)
Institute
- Institut für Physik und Astronomie (6)
- Institut für Geowissenschaften (4)
- Department Psychologie (3)
- Institut für Biochemie und Biologie (3)
- Institut für Ernährungswissenschaft (2)
- Department Erziehungswissenschaft (1)
- Institut für Umweltwissenschaften und Geographie (1)
- Wirtschaftswissenschaften (1)
There is no consensus on which statistical model estimates school value-added (VA) most accurately. To date, the two most common statistical models used for the calculation of VA scores are two classical methods: linear regression and multilevel models. These models have the advantage of being relatively transparent and thus understandable for most researchers and practitioners. However, these statistical models are bound to certain assumptions (e.g., linearity) that might limit their prediction accuracy. Machine learning methods, which have yielded spectacular results in numerous fields, may be a valuable alternative to these classical models. Although big data is not new in general, it is relatively new in the realm of social sciences and education. New types of data require new data analytical approaches. Such techniques have already evolved in fields with a long tradition in crunching big data (e.g., gene technology). The objective of the present paper is to competently apply these "imported" techniques to education data, more precisely VA scores, and assess when and how they can extend or replace the classical psychometrics toolbox. The different models include linear and non-linear methods and extend classical models with the most commonly used machine learning methods (i.e., random forest, neural networks, support vector machines, and boosting). We used representative data of 3,026 students in 153 schools who took part in the standardized achievement tests of the Luxembourg School Monitoring Program in grades 1 and 3. Multilevel models outperformed classical linear and polynomial regressions, as well as different machine learning models. However, it could be observed that across all schools, school VA scores from different model types correlated highly. Yet, the percentage of disagreements as compared to multilevel models was not trivial and real-life implications for individual schools may still be dramatic depending on the model type used. Implications of these results and possible ethical concerns regarding the use of machine learning methods for decision-making in education are discussed.
Nonstationary coherence-incoherence patterns in nonlocally coupled heterogeneous phase oscillators
(2020)
We consider a large ring of nonlocally coupled phase oscillators and show that apart from stationary chimera states, this system also supports nonstationary coherence-incoherence patterns (CIPs). For identical oscillators, these CIPs behave as breathing chimera states and are found in a relatively small parameter region only. It turns out that the stability region of these states enlarges dramatically if a certain amount of spatially uniform heterogeneity (e.g., Lorentzian distribution of natural frequencies) is introduced in the system. In this case, nonstationary CIPs can be studied as stable quasiperiodic solutions of a corresponding mean-field equation, formally describing the infinite system limit. Carrying out direct numerical simulations of the mean-field equation, we find different types of nonstationary CIPs with pulsing and/or alternating chimera-like behavior. Moreover, we reveal a complex bifurcation scenario underlying the transformation of these CIPs into each other. These theoretical predictions are confirmed by numerical simulations of the original coupled oscillator system.
While previous research underscores the role of leaders in stimulating employee voice behaviour, comparatively little is known about what affects leaders' support for such constructive but potentially threatening employee behaviours. We introduce leader member exchange quality (LMX) as a central predictor of leaders' support for employees' ideas for constructive change. Apart from a general benefit of high LMX for leaders' idea support, we propose that high LMX is particularly critical to leaders' idea support if the idea voiced by an employee constitutes a power threat to the leader. We investigate leaders' attribution of prosocial and egoistic employee intentions as mediators of these effects. Hypotheses were tested in a quasi-experimental vignette study (N = 160), in which leaders evaluated a simulated employee idea, and a field study (N = 133), in which leaders evaluated an idea that had been voiced to them at work. Results show an indirect effect of LMX on leaders' idea support via attributed prosocial intentions but not via attributed egoistic intentions, and a buffering effect of high LMX on the negative effect of power threat on leaders' idea support. Results differed across studies with regard to the main effect of LMX on idea support.
After the United Kingdom has left the European Union it remains unclear whether the two parties can successfully negotiate and sign a trade agreement within the transition period. Ongoing negotiations, practical obstacles and resulting uncertainties make it highly unlikely that economic actors would be fully prepared to a “no-trade-deal” situation. Here we provide an economic shock simulation of the immediate aftermath of such a post-Brexit no-trade-deal scenario by computing the time evolution of more than 1.8 million interactions between more than 6,600 economic actors in the global trade network. We find an abrupt decline in the number of goods produced in the UK and the EU. This sudden output reduction is caused by drops in demand as customers on the respective other side of the Channel incorporate the new trade restriction into their decision-making. As a response, producers reduce prices in order to stimulate demand elsewhere. In the short term consumers benefit from lower prices but production value decreases with potentially severe socio-economic consequences in the longer term.
The Cluster mission has produced a large data set of electron flux measurements in the Earth's magnetosphere since its launch in late 2000. Electron fluxes are measured using Research with Adaptive Particle Imaging Detector (RAPID)/Imaging Electron Spectrometer (IES) detector as a function of energy, pitch angle, spacecraft position, and time. However, no adiabatic invariants have been calculated for Cluster so far. In this paper we present a step-by-step guide to calculations of adiabatic invariants and conversion of the electron flux to phase space density (PSD) in these coordinates. The electron flux is measured in two RAPID/IES energy channels providing pitch angle distribution at energies 39.2-50.5 and 68.1-94.5 keV in nominal mode since 2004. A fitting method allows to expand the conversion of the differential fluxes to the range from 40 to 150 keV. Best data coverage for phase space density in adiabatic invariant coordinates can be obtained for values of second adiabatic invariant, K, similar to 10(2), and values of the first adiabatic invariant mu in the range approximate to 5-20 MeV/G. Furthermore, we describe the production of a new data product "LSTAR," equivalent to the third adiabatic invariant, available through the Cluster Science Archive for years 2001-2018 with 1-min resolution. The produced data set adds to the availability of observations in Earth's radiation belts region and can be used for long-term statistical purposes.
In older persons, the origin of malnutrition is often multifactorial with a multitude of factors involved. Presently, a common understanding about potential causes and their mode of action is lacking, and a consensus on the theoretical framework on the etiology of malnutrition does not exist. Within the European Knowledge Hub "Malnutrition in the Elderly (MaNuEL)," a model of "Determinants of Malnutrition in Aged Persons" (DoMAP) was developed in a multistage consensus process with live meetings and written feedback (modified Delphi process) by a multiprofessional group of 33 experts in geriatric nutrition. DoMAP consists of three triangle-shaped levels with malnutrition in the center, surrounded by the three principal conditions through which malnutrition develops in the innermost level: low intake, high requirements, and impaired nutrient bioavailability. The middle level consists of factors directly causing one of these conditions, and the outermost level contains factors indirectly causing one of the three conditions through the direct factors. The DoMAP model may contribute to a common understanding about the multitude of factors involved in the etiology of malnutrition, and about potential causative mechanisms. It may serve as basis for future research and may also be helpful in clinical routine to identify persons at increased risk of malnutrition.
During reading, rapid eye movements (saccades) shift the reader's line of sight from one word to another for high-acuity visual information processing. While experimental data and theoretical models show that readers aim at word centers, the eye-movement (oculomotor) accuracy is low compared to other tasks. As a consequence, distributions of saccadic landing positions indicate large (i) random errors and (ii) systematic over- and undershoot of word centers, which additionally depend on saccade lengths (McConkie et al.Visual Research, 28(10), 1107-1118,1988). Here we show that both error components can be simultaneously reduced by reading texts from right to left in German language (N= 32). We used our experimental data to test a Bayesian model of saccade planning. First, experimental data are consistent with the model. Second, the model makes specific predictions of the effects of the precision of prior and (sensory) likelihood. Our results suggest that it is a more precise sensory likelihood that can explain the reduction of both random and systematic error components.
A large part of classical visual psychophysics was concerned with the fundamental question of how pattern information is initially encoded in the human visual system. From these studies a relatively standard model of early spatial vision emerged, based on spatial frequency and orientation-specific channels followed by an accelerating nonlinearity and divisive normalization: contrast gain-control. Here we implement such a model in an image-computable way, allowing it to take arbitrary luminance images as input. Testing our implementation on classical psychophysical data, we find that it explains contrast detection data including the ModelFest data, contrast discrimination data, and oblique masking data, using a single set of parameters. Leveraging the advantage of an image-computable model, we test our model against a recent dataset using natural images as masks. We find that the model explains these data reasonably well, too. To explain data obtained at different presentation durations, our model requires different parameters to achieve an acceptable fit. In addition, we show that contrast gain-control with the fitted parameters results in a very sparse encoding of luminance information, in line with notions from efficient coding. Translating the standard early spatial vision model to be image-computable resulted in two further insights: First, the nonlinear processing requires a denser sampling of spatial frequency and orientation than optimal coding suggests. Second, the normalization needs to be fairly local in space to fit the data obtained with natural image masks. Finally, our image-computable model can serve as tool in future quantitative analyses: It allows optimized stimuli to be used to test the model and variants of it, with potential applications as an image-quality metric. In addition, it may serve as a building block for models of higher level processing.
Trade plays a key role in the spread of alien species and has arguably contributed to the recent enormous acceleration of biological invasions, thus homogenizing biotas worldwide. Combining data on 60-year trends of bilateral trade, as well as on biodiversity and climate, we modeled the global spread of plant species among 147 countries. The model results were compared with a recently compiled unique global data set on numbers of naturalized alien vascular plant species representing the most comprehensive collection of naturalized plant distributions currently available. The model identifies major source regions, introduction routes, and hot spots of plant invasions that agree well with observed naturalized plant numbers. In contrast to common knowledge, we show that the 'imperialist dogma,' stating that Europe has been a net exporter of naturalized plants since colonial times, does not hold for the past 60 years, when more naturalized plants were being imported to than exported from Europe. Our results highlight that the current distribution of naturalized plants is best predicted by socioeconomic activities 20 years ago. We took advantage of the observed time lag and used trade developments until recent times to predict naturalized plant trajectories for the next two decades. This shows that particularly strong increases in naturalized plant numbers are expected in the next 20 years for emerging economies in megadiverse regions. The interaction with predicted future climate change will increase invasions in northern temperate countries and reduce them in tropical and (sub) tropical regions, yet not by enough to cancel out the trade-related increase.
The Sea-level Response to Ice Sheet Evolution (SeaRISE) effort explores the sensitivity of the current generation of ice sheet models to external forcing to gain insight into the potential future contribution to sea level from the Greenland and Antarctic ice sheets. All participating models simulated the ice sheet response to three types of external forcings: a change in oceanic condition, a warmer atmospheric environment, and enhanced basal lubrication. Here an analysis of the spatial response of the Greenland ice sheet is presented, and the impact of model physics and spin-up on the projections is explored. Although the modeled responses are not always homogeneous, consistent spatial trends emerge from the ensemble analysis, indicating distinct vulnerabilities of the Greenland ice sheet. There are clear response patterns associated with each forcing, and a similar mass loss at the full ice sheet scale will result in different mass losses at the regional scale, as well as distinct thickness changes over the ice sheet. All forcings lead to an increased mass loss for the coming centuries, with increased basal lubrication and warmer ocean conditions affecting mainly outlet glaciers, while the impacts of atmospheric forcings affect the whole ice sheet.