Search

Background/Context: There is continuing debate among social scientists and educators about the role of school-to-school differences in generating educational inequality. Are some students high achieving because they attend School A, while others struggle because they attend School B, as critical discourse on schools argues? Alternatively, is educational inequality driven largely by social forces outside of the school, in the home and neighborhood environment, or by educational processes that are largely common across schools as much social science research argues? Analyses of school achievement, and in particular test score gains from year-to-year, suggest very small between-school differences. Yet, analyses of test score data alone may fail to reveal important school-to-school differences that affect the quality of the classroom experience and a variety of educational outcomes. Purpose/Objective: We provide evidence on the following research questions. What is the magnitude of school-to-school variation in instructional practice, as captured by multiple measures? Are some domains of instruction (e.g. behavioral management) more variable between-schools than others? To what extent are school-to-school differences in instruction associated with compositional characteristics of students and teachers? Research Design: This study relies on the Measures of Effective Teaching Study data, which offer an unprecedented set of observations of teachers' instruction scored on state of the art observational protocols. To examine the extent of school-to-school variation in instructional practice in elementary and middle schools, we conducted a decomposition of variance analysis using summary scores on multiple measures. We further examine behavioral climate as revealed during instruction separately from overall instructional practice. Next, we examine differences in instruction associated with compositional characteristics of students using multilevel models. Finally, we utilize an innovative two-stage statistical adjustment strategy to more narrowly identify the possible association between composition and teaching practice due to school-to-school teacher sorting. Findings/Results: The basic descriptive results from this study suggest a middle-view of schoolto-school differences in instruction. We find that substantial school-level variation in instruction exists, with 30% or more of the total variance in instruction lying between schools in these data. Behavioral climate during instruction appears to be particularly salient, and especially in elementary schools. Much of the between school variance we identify, in some cases 40% or more, is readily explained by simple measures of socio-demographic composition, including in particular the racial make-up of schools in the MET districts. Finally, some evidence from a statistical adjustment method suggests that teacher sorting, rather than measurement bias and teacher adaptation is principally responsible for school-to-school differences in instruction. Conclusions/Recommendations: More than an academic debate, basic differences between schools in the quality of the learning environment, along with parental understandings and beliefs about school effects, are potentially important drivers of school and neighborhood sorting and segregation, and even public investment in schooling. Additionally, this question carries continued policy relevance as states adopt and revise teacher and school accountability frameworks that implicitly attribute school-to-school differences to organizational functioning, and seek to carry out instructional improvement efforts in targeted schools. The basic descriptive results from this study suggest school-level differences are not as great as suggested by critical theory and the public discourse, but nor are they as inconsequential as one might infer from some social science research or the literature on value-added differences between schools. [This paper will be published in "Teachers College Record."]

The remnant of SN 1987A is the best-studied object of its kind. The rich data-set of its thermal and non-thermal emission across the electromagnetic spectrum poses a unique testbed for the elaboration of particle-acceleration theory. We use 2D simulations of the progenitor's wind to obtain hydro-profiles for the medium around the supernova explosion. Various cones along prominent features of the ambient medium are then used in our time-dependent acceleration code RATPaC to model the evolution of the emission of SN 1987A and compare it to observational data. We solve for the transport of cosmic rays and the hydrodynamical flow, in the test-particle limit.The simulation code relies on 1D profiles but the large expansion speed of the young remnant renders lateral transport unimportant. We find that the increase in thermal X-ray emission predates the increase in the low-energy gamma-ray brightness by several years. The increase of the gamma-ray brightness at lower energies is followed by a smooth increase at the highest energies. The gamma-ray spectrum at the highest energies appears soft during the brightening but hardens as more material in the equatorial ring gets shocked. The X-ray and gamma-ray brightness remain almost constant once the SNR blast-wave passed the region of peak-density in the equatorial plane., Comment: 8 pages, ICRC2021 proceedings

We examine the ability of supervised text classification models to identify several discourse properties from teachers' speech with an eye for providing teachers with meaningful automated feedback about the quality of their classroom discourse. We collected audio recordings from 28 teachers from 10 schools in 164 authentic classroom sessions, which we then automatically transcribed into text utterances and then manually coded to identify whether: (1) the utterance contained a question (as opposed to a statement), (2) the question or statement was Instructional vs. Non-Instructional, and (3) the question or statement was Content-Specific. We experimented with Random Forest classifiers and engineered (linguistic, acoustic-prosodic, and contextual) features vs. open-vocabulary n-grams as features to discriminate these discourse variables at the utterance level in a teacher-independent fashion. We achieved AUC scores ranging from 0.71 to 0.77 using open-vocabulary language modeling, which were well above chance (AUC = 0.5), an important step towards our predominant goal of constructing of an automated feedback system for teacher reflection and learning. [For the full proceedings, see ED599096.]

This paper studies how the effectiveness of teachers varies by classroom composition. We combine random assignment of teachers to classrooms with rich measures of teacher effectiveness based on a popular observational protocol, Framework for Teaching, to overcome key endogeneity concerns related to measurement and matching. We find that complementarities between classroom composition and teaching practice play a significant role in student achievement. We identify two main mechanisms that are driving this result: (1) negative interactions between challenging and/or student-centered practices and heterogeneity in classroom prior achievement, and (2) positive interactions between classroom management skills and average classroom prior achievement. Our findings illustrate the multidimensional nature of teacher effectiveness and have important implications for prescribing teaching practice and evaluating teachers. Simulations show that teacher rankings change substantially simply from within-school classroom reallocations, suggesting the need for caution when using popular teaching evaluation rubrics in high-stakes settings.

Personalized learning environments requiring the elicitation of a student's knowledge state have inspired researchers to propose distinct models to understand that knowledge state. Recently, the spotlight has shone on comparisons between traditional, interpretable models such as Bayesian Knowledge Tracing (BKT) and complex, opaque neural network models such as Deep Knowledge Tracing (DKT). Although DKT appears to be a powerful predictive model, little effort has been expended to dissect the source of its strength. We begin with the observation that DKT differs from BKT along three dimensions: (1) DKT is a neural network with many free parameters, whereas BKT is a probabilistic model with few free parameters; (2) a single instance of DKT is used to model all skills in a domain, whereas a separate instance of BKT is constructed for each skill; and (3) the input to DKT interlaces practice from multiple skills, whereas the input to BKT is separated by skill. We tease apart these three dimensions by constructing versions of DKT which are trained on single skills and which are trained on sequences separated by skill. Exploration of three data sets reveals that dimensions (1) and (3) are critical; dimension (2) is not. Our investigation gives us insight into the structural regularities in the data that DKT is able to exploit but that BKT cannot. [For the full proceedings, see ED593090.]