Andre A. Rupp 1230A Benjamin Building University of Maryland College Park, MD 20742 ruppandr@umd.edu (301) 405-3623 Voice (301) 314-9245 Fax

I initially studied in Hamburg, Germany, to become a secondary school teacher for English, French, and Mathematics. Through completing my Master’s work in applied linguistics and statistics in the U.S. and my Ph.D. work in research methods in Canada, I got lured into the beauty of academic research and teaching. I started my academic career at the bilingual University of Ottawa in Canada. Until recently, I worked at the Institute for Educational Progress (IQB) in Berlin, Germany, in an interdisciplinary team on developing national standards-based assessments for English as a first foreign language.

My current research interests center around cognitively-grounded assessment approaches and associated statistical models. Themes in my research that I would like to pursue in more depth in the future include investigating the theoretical potential and practical limitations of cognitive diagnosis models, developing principled diagnostic assessment approaches for complex skill sets using innovative digital technologies, and researching the impact of large-scale design structures on model fit within such contexts. Some of my related work has focused on investigating parameter invariance in latent variable models, where my most recent attention has been on the influence of missing data on the effectiveness of methods for investigating differential item functioning.

I continue to be involved in two projects with the Institute for Educational Progress (Insitut zur Qualitätsentwicklung im Bildungswesen, IQB) at the Humboldt University in Berlin, Germany (www.iqb.hu-berlin.de), which is the national assessment institute charged with developing standards-based assessments for a variety of subjects in lower secondary school and elementary school. The first project is conducted in collaboration with Dr. Oliver Wilhelm at the Humboldt University in Berlin whereas the second project is conducted in collaboration with Dr. Hans Anand Pant at the ISQ in Berlin (www.isq-bb.de), Dr. Olaf Köller at the IQB, and the BEAR research group at the University of California at Berkeley (bearcenter.berkeley.edu)

The first project is concerned with investigating the consistency of score profiles for large-scale diagnostic assessments that are derived from multidimensional models in item response theory as well as multidimensional cognitive diagnosis models. In this project, children respond to a newly developed diagnostic assessment for basic arithmetic ability at the elementary level as well as several instruments that tap basic mental capacities that are believed to influence their response processes. Through comprehensive psychometric analyses, a thorough investigation of the degree to which fine-grained inferences about cognitive response processes for this diagnostic test can be justified is conducted. This provides insight into the conditions under which the psychological multidimensionality of constructs can be empirically captured via multidimensional scaling models so that instructionally relevant decisions for remediation and support can be reliably supported.

The second project focuses on investigating how two different methods of setting cut scores and two different panel compositions, which are experimentally varied, influence the consistency of cut-scores on unidimensional proficiency scales for reading comprehension, listening comprehension, and written expression in the context of a large-scale standards-based assessment. Furthermore, the project investigates how experts weight and synthesize various pieces of empirical evidence regarding the potential locations of cut scores. These pieces of evidence come from psychometric analyses of item difficulty features, unidimensional scaling models, consensual standard-setting procedures, and descriptions of cognitive response processes. Furthermore, the project seeks to understand how information about the setting of cut-scores can be most effectively communicated so that the chances for misinterpretations of the resulting proficiency levels in the media can be minimized.