New perspectives on social and psychological processes (18)

Chair: Yasuo Miyazaki, Tuesday 21st July, 15.00 - 16.20, Dirac Room, Fisher Building. 

Kian Abolfazlian, Spider: Business Ideas Architects, Inc., Aarhus, Denmark. How to quantitatively measure the psychological phenomena of team synergy in an organizational setting. (231)

Jan-Marten Ihme, Leibniz Institute for Science Education, Kiel, Germany. Construction of figural matrices tasks with predicted difficulty. (183)

Han van der Maas, Sven Stringer, Denny Borsboom and Gunter Maris, Department of Psychology, University of Amsterdam, The Netherlands. The derivation of item response models from sequential sampling models of choice. (195)

Teresa Calapez and Madalena Ramos, DMQ – Quantitative Methods Department and UNIDE/STATMATH,  Lisbon University Institute, Portugal. Do different presentations of Likert-type items lead to differences in structure?: A field study using linear and nonlinear PCA. (139)

 

ABSTRACTS

How to quantitatively measure the psychological phenomena of team synergy in an organizational setting. (231)
Kian Abolfazlian
In an organizational setting, especially at the group/team level, it is of outmost importance to be able to measure quantitatively and predict the achievable team synergy, when deciding how to allocate different resources needed to perform a specific task. Most importantly, it is paramount to have a clear idea of which professionals to choose in order to achieve the adequate synergy level for the task at hand. Furthermore it is of equal importance to measure the changes in the synergy level of an existing team. The change may be caused by either the introduction of new members or the alteration of the organizational control mechanisms, such as roles, rules and instruments, which affect the working of the team. Here, we treat the team synergy as a psychological phenomenon and independent of the usual economical post-measurement models. In this paper, we, firstly, introduce algorithms to turn the organizational control mechanisms to measurable spaces, and secondly, we entertain an algebraic model to measure synergy level of teams.

Construction of figural matrices tasks with predicted difficulty. (183)
Jan-Marten Ihme
Raven Matrices Tasks are considered to be the best marker for fluid intelligence. They are suitable for rule-based construction because of their clearly describable structure. The difficulty of a task should be determined by its structure, while the empirical correlation often is not higher than .40. This might be the case since the item description in the model often is incomplete, because only rules, drawing principals and directions are coded, but not the figures and elements used in the specific item. In this study, matrices tasks were constructed so that all structure attributes were set constant except for the rules, so that every possible cognitive demand is either constant or controlled. An online test was carried out. An analysis using the Linear Logistic Test Model (LLTM) revealed that the rules clear up about 95 percent of variance of the item difficulties. Preliminary results of a second study show that the estimated parameters can be used for a prediction of item difficulties of new items constructed be the same rationale.

The derivation of item response models from sequential sampling models of choice. (195)
Han van der Maas, Sven Stringer, Denny Borsboom and Gunter Maris
We discuss the work of Tuerlinckx en De Boeck (2005), who derived the popular two parameter logistic model (2PLM) for item responses from the diffusion model, a stochastic sequential sampling model for simple two choice decisions. Other derivations of IRT models are based on desirable statistical or measurement properties but have no bearing on the processes that generate the data. However, we contend that the derivation of Tuerlinckx en De Boeck is only valid when the 2PLM is used for items with two response options. We extend their work by presenting a formal extension of the 2PLM and the diffusion model to allow for more than two choices (MC2PLM). The behavior of this simple multiple choice diffusion model is consistent with Hick’s law. Another limitation of Tuerlinckx en De Boeck’s derivation is that it does handle guessing by subjects for whom the item is too difficult. This frequently occurs in multiple choice high stake ability testing as in exams. We propose a new restricted 2PLM that handles guessing in IRT. This new model implies new interpretations of the standard parameters of IRT models.

Do different presentations of Likert-type items lead to differences in structure?: A field study using linear and nonlinear PCA. (139)
Teresa Calapez and Madalena Ramos
Although methodologically Likert-type variables are often classified as ordinal, to use them as numerical is common, e.g. when carrying out a dimensionality reduction by a Principal Components Analysis, which decomposes a correlation matrix. Nevertheless, a valid alternative to PCA for ordinal variables exists: nonlinear PCA, which, under the acronym CATPCA, was implemented within the commercial package SPSS. Our objective is to study in what measure different presentations of Likert-type items (including "continuous" options, i.e. marking the option on a straight line, with or without middle point and the use of all anchors vs. extreme-only labels) cause differences in structure when subjected to linear and nonlinear data reduction techniques. Participants were first year students of several courses. There were four types of questionnaires: continuous, extreme-labelled scale, with a middle mark (A); continuous extreme-labelled scale, with no middle mark (B); 5-point all-labelled Likert-type items (C); and 5-point extreme only labelled Likert-type items (D). Questionnaires were affected systematically in each class, so that an approximate number of each type was obtained. Data in each group will be subjected to both PCA and CATPCA. Structures will be compared using either canonical correlations (within the same group) or Tucker congruency coefficients (within and between groups). References: Meulman, Jacqueline; Van der Kooij, Anita; Heiser, Willem (2004). "Principal Components Analysis with Nonlinear Optimal Scaling Transformations for Ordinal and Nominal Data", in The Sage Handbook of Quantitative Methodology for the Social Sciences, David Kaplan (Ed.) 49-70, Sage. Linting, Mariëlle; Meulman, Jacqueline; Groenen, Patrick; Van der Kooij, Anita (2007). "Nonlinear Principal Components Analysis: Introduction and Application", Psychological Methods, 12(3), 336-358. Heiser, Willem; Meulman, Jacqueline (2005). SPSS Categories 14.0, Chicago, SPSS Inc. Göb, Rainer; McCollin, Christopher; Ramalhoto, Fernanda (2007). "Ordinal Methodology in the Analysis of Likert Scales", Quality & Quantity, 41, 601-626. Lorenzo-Seva, Urbano; ten Berge, Jos (2006). "Tucker's congruence coefficient as a meaningful index of factor similarity", Methodology 2 (2), 57-64.