David Andrich is Chapple Professor, Graduate School of Education, The University of Western Australia.
He obtained a bachelor degree in Mathematics and his Masters degree in Education from The University of Western Australia and his PhD from the University of Chicago, for which he was awarded the Susan Colver Rosenberger prize for the best research thesis in the Division of the Social Sciences. He returned to The University of Western Australia, and in 1985 was appointed Professor of Education at Murdoch University, also in Western Australia. In 2007 he returned to The University of Western Australia as the Chapple Professor of Education. In 1977 he spent 6 months as a Research Fellow at the Danish Institute for Educational Research working with Georg Rasch and he has been a Visiting Professor at the University of Trento in Italy for two periods. He has held major research grants from the Australian Research Council continuously since 1985 and has conducted commissioned government research at both the national and state levels. In 1990, he was elected Fellow of the Academy of Social Sciences of Australia for his contributions to measurement in the social sciences. He is especially known for his work in modern test theory, and in particular Rasch models for measurement, ranging in topics from the philosophy of measurement, through model exposition and interpretation, to software development. He has published in Educational, Psychological, Sociological and Statistical journals. He is the author of Rasch Models for Measurement (Sage) and coauthor of the software package Rasch Unidimensional Measurement Models (RUMMLab).
David Andrich’s current research in applying Rasch models for measurement is has two strands. The first involves articulating a research and assessment paradigm that is different from the traditional in which statistical models are applied. In the traditional paradigm, the case for choosing any model to summarise data is that it fits the data at hand; in contrast, in applying the paradigm of Rasch models, the case for these models is that if the data fit the model, then, within a frame of reference, they provide invariance of comparisons of persons with respect to items, and vice versa. Then any misfit between the data and the chosen Rasch model is seen as an anomaly that needs to be explained by qualitatively by reference to the theory behind the construction of the instrument, and the operational aspects of its application. He argues that this approach improves the quality of social measurement, including in education, psychology, sociology, economics and in health outcomes. The second area of research is further articulating the implications of the Rasch models and development of complementary software, to better understand a range of anomalies, for example, how to identify guessing in multiple choice items, how to identify and handle response dependence between items, and mutldimensionality. He has also recently published the paper which shows how person location estimates can be obtained independently of all test parameters using the general unidimensional Rasch model in the case where each person has sat a multiple of tests, for example for selection for university entry. Andrich, D. (2010) Sufficiency and conditional estimation of person parameters in the polytomous Rasch model. Psychometrika. (Online First Publication).