Suchergebnisse

This book provides a clear introduction to this important area of statistics. The author provides a wide of coverage of different kinds of multilevel models, and how to interpret different statistical methodologies and algorithms applied to such models. This 4th edition reflects the growth and interest in this area and is updated to include new chapters on multilevel models with mixed response types, smoothing and multilevel data, models with correlated random effects and modeling with variance

This book provides a clear introduction to this important area of statistics. The author provides a wide of coverage of different kinds of multilevel models, and how to interpret different statistical methodologies and algorithms applied to such models. This 4th edition reflects the growth and interest in this area and is updated to include new chapters on multilevel models with mixed response types, smoothing and multilevel data, models with correlated random effects and modeling with variance.

Random cross-classifications of units can arise at any level of a data hierarchy. For example, school students may be classified both by the schools they attend and their neighborhoods of residence. This article explores the issues of efficiently modeling such data and gives an example from a study of parental choice of schools.

Since 1992, the UK Government has published so-called 'school league tables' summarizing the average General Certificate of Secondary Education (GCSE) 'attainment' and 'progress' made by pupils in each state-funded secondary school in England. While the headline measure of school attainment has remained the percentage of pupils achieving five or more good GCSEs, the headline measure of school progress has changed from 'value-added' (2002-2005) to 'contextual value-added' (2006-2010) to 'expected progress' (2011-2015) to 'progress 8' (2016-). This paper charts this evolution with a critical eye. First, we question the Government's justifications for scrapping contextual value-added. Second, we argue that the current expected progress measure suffers from fundamental design flaws. Third, we show that the differences between expected progress and contextual value added are considerable leading to fundamentally different school rankings. Fourth, we discuss how 'progress 8' attempts to address some, but not all, of the flaws in expected progress. We conclude that all these progress measures and school league tables more generally should be viewed with far more scepticism and interpreted far more cautiously than they have often been to date.

Since 1992, the UK Government has published so-called 'school league tables' summarizing the average General Certificate of Secondary Education (GCSE) 'attainment' and 'progress' made by pupils in each state-funded secondary school in England. While the headline measure of school attainment has remained the percentage of pupils achieving five or more good GCSEs, the headline measure of school progress has changed from 'value-added' (2002-2005) to 'contextual value-added' (2006-2010) to 'expected progress' (2011-2015) to 'progress 8' (2016-). This paper charts this evolution with a critical eye. First, we question the Government's justifications for scrapping contextual value-added. Second, we argue that the current expected progress measure suffers from fundamental design flaws. Third, we show that the differences between expected progress and contextual value added are considerable leading to fundamentally different school rankings. Fourth, we discuss how 'progress 8' attempts to address some, but not all, of the flaws in expected progress. We conclude that all these progress measures and school league tables more generally should be viewed with far more scepticism and interpreted far more cautiously than they have often been to date.

The authors present a factor model for the analysis of categorical responses with a two-level hierarchical structure. The model allows for multiple, potentially correlated factors at each level as well as covariate effects on the responses. Estimation using Markov chain Monte Carlo (MCMC) methods is described. The methodology is applied in an analysis of women's status in Bangladesh. Two dimensions of women's status are considered—social independence and decision-making power—and the factor structure of the responses at the woman and district levels is explored.