Anton Formann


Anton K. Formann was an Austrian research psychologist, statistician, and psychometrician. He is renowned for his contributions to item response theory, latent class analysis, the measurement of change, mixture models, categorical data analysis, and quantitative methods for research synthesis.

Biography

Anton K. Formann studied psychology with statistics and anthropology at the University of Vienna, Austria, where he received his PhD in psychology in 1973 under the supervision of :de:Gerhard H. Fischer|Gerhard H. Fischer at the university's Department of Psychology. He worked as a post doc researcher and Assistant Professor at Fischer's division until 1985, when he earned his postdoctoral professorial qualification and became Associate Professor at the University of Vienna.
He also studied statistics at Sheffield Hallam University where he graduated in 1998. In 1999, he gained his second postdoctoral professional qualification.
In 2004, after being substitute chair holder for 5 years, he became full professor for psychological methods at the University of Vienna, succeeding the chair of mathematical psychology of :de:Gerhard H. Fischer|Gerhard H. Fischer.
From 2005 onwards, Formann was Vice Head of the Department of Basic Psychological Research within the Faculty of Psychology at the University of Vienna, and during 2006-08 additionally Vice Dean of the Faculty.

Scientific Work

Formann led long-standing research collaborations with colleagues in the statistical, medical, and psychological sciences. His substantial research activities in all these fields are documented in numerous books and more than 50 publications in prestigious high-impact journals, including Biometrics, the Journal of the American Statistical Association, the British Journal of Mathematical and Statistical Psychology, and Psychometrika.

Item response theory (Rasch models)

Formann was one of the first researchers who documented problems with Rasch model tests, in particular with Andersen's likelihood-ratio test which arise under certain conditions if it is employed conventionally. As a senior author, Formann also showed that the common assumption that the EM estimation of the two-parameter logistic model is not influenced by initial values is incorrect.

Viennese Matrices Test

Formann was probably the first researcher to practically apply Fischer's linear logistic test model for test development. The LLTM is a special case of the Rasch model, which allows the construction of items with item difficulties based on the user's demand. This resulted in the development of a Rasch-scaled abstract reasoning test which has since been widely used in research and practice. A revised version of this language-free intelligence test that has been calibrated against large contemporary samples of men and women is forthcoming.

Latent Class Analysis

For his first habilitation, Formann published a comprehensive monograph on latent class analysis which continues to be widely cited for its clarity, depth, and originality, and hence is considered a true modern classic on this topic.

Quantitative Methods for Research Synthesis (Meta-Analysis)

In his later research, Formann addressed, among other things, the problem of publication bias in meta-analytic research. He introduced a novel method that allows estimating the proportion of studies missing in meta-analysis due to publication bias based on the truncated normal distribution.
In 2010, as the senior author, Formann debunked in a meta-analysis the famous Mozart effect as a myth.

Other

Newcomb-Benford Law

Formann provided an alternative explanation for the Newcomb-Benford law – a formalisation of the remarkable observation that the frequencies with which the leading digits of numbers occur in large data sets are far away from being uniform. In addition to the prevailing explanations based on scale- and base invariance, Formann directed the attention to the interrelation between the distribution of the significant digits and the distribution of the observed variable. He showed in a simulation study that long right-tailed distributions of a random variable are compatible with the Newcomb-Benford law, and that for distributions of the ratio of two random variables the fit generally improves.

Piaget's Water Level Task

The water-level task refers to a task developed by Jean Piaget where bottles filled with different levels of water are presented in different angles of orientation. It is used to assess the level of mental development of spatial abilities. Formann criticized the established method of dichotomizing water-level responses by the subjects into "right" versus "wrong" – he showed that this method was inappropriate because it ignored the heterogeneity of the task difficulties - and instead recommended using latent class models or Rasch models. He showed that subjects and tasks can be arrayed on a unidimensional scale and, by employing the linear logistic test model, that the task difficulty could be attributed to a single parameter associated with the angle of inclination of the bottle. Furthermore, he provided the first empirical data of task performance of the elderly and found that there was an age-associated non-linear decline of performance.

Misconception of Probability

Formann compared the performance in the classic birthday problem and the birthmate problem in psychology undergraduates, casino visitors, and casino employees. Psychology students and women did better on both task types, but were less confident about their estimates than casino visitors or personnel and men. Higher confidence ratings were related to subjective estimates that were closer to the solutions of birthday problems, but not of birthmate problems.

Parallel Analysis in Retrieving Unidimensionality in the Presence of Binary Data

Formann provided both theoretical and empirical evidence that the application of the parallel analysis for uncovering the factorial structure of binary variables is not appropriate. Results of a simulation study showed that sample size, item discrimination, and type of correlation coefficient considerably influence the performance of parallel analysis.

Selected publications

Papers