CONTENTS
A reasonable approach to check a psychological test’s long ago standardization – applied for the Adaptive Intelligence Diagnosticum (AID 3)
Klaus D. Kubinger & Thomas Suster
.PDF of the full article
Visualizing Rasch item fit using conditional item characteristic curves in R
Ann-Sophie Buchardt, Karl Bang Christensen & Sidsel Normann Jensen
.PDF of the full article
Detecting and Differentiating Extreme and Midpoint Response Styles in Rating-type scale using Tree-Based Item Response Models: Simulation Study and Empirical Evidence
Artur Pokropek, Lale Khorramdel & Matthias von Davier
.PDF of the full article
Extending GMX: Conditional Likelihood Ratio Test and Extended Graphical Model Checks with psychotools
Rainer W. Alexandrowicz
.PDF of the full article
The Need for Quantification
Theoretical considerations of a construct and evaluation of a scale for its measurement
Henrik Gast, Thomas Ostermann, Tugba Kapanci & Stefan J. Troche
.PDF of the full article
Breadth of Data Dispersion and Number of Variables as Sources of Measurement Effects on Factor Variances in Confirmatory Factor Analysis
Karl Schweizer, Stefan Troche & Tengfei Wang
.PDF of the full article
A reasonable approach to check a psychological test’s long ago standardization – applied for the Adaptive Intelligence Diagnosticum (AID 3)
Klaus D. Kubinger & Thomas Suster
Abstract
Due to the DIN 33430 [Requirements on procedures for the assessment of professional aptitude], a psychological test’s standardization must be verified every eight years. However, in particular for tests that are to be individually administrated, standardization means a vast and very expensive undertaking. Therefore, sampling new data only for checking a psychological test’s long ago standardization is to be minimized or preferably avoided. This paper suggests to use data of case reports from pertinent institutions instead. It is outlined that this approach is not only economical but also very reliable – although such data is most likely not representative with regard to the population in question: By this means easily not only a single but rather several surveys are possible within the time line from the original standardization data sampling to the actual point in time; they disclose whether some linear or at least some monotonically decreasing or increasing progression of the mean of test-scores occurs or just random fluctuations of the test-score level take place. Even if the results force the revision of the test’s standardization, this approach is probably of further use; regression analysis could lead to a predicted value of the test-score’s mean and standard deviation in the next calendar year(s), which can be the basis for a re-standardization. Suitability of the current suggested approach is illustrated through an example that concerns a single subtest of the intelligence test-battery Adaptive Intelligence Diagnosticum (AID 3; Kubinger & Holocher-Ertl, 2014).
Keywords: test standardization; representative sampling; DIN 33430; two-way analysis of variance (cross-classification, mixed model); Adaptive Intelligence Diagnosticum (AID 3)
Klaus D. Kubinger, PhD.
Professorial Research Fellow
University of Vienna, Faculty of Psychology
Liebiggasse 5
1010 Vienna, Austria
email: klaus.kubinger@univie.ac.at
Visualizing Rasch item fit using conditional item characteristic curves in R
Ann-Sophie Buchardt, Karl Bang Christensen, and Sidsel Normann Jensen
Abstract
New R computer routines that support rigorous statistical validation of psychological tests have recently appeared. We illustrate how Rasch item fit can be evaluated visually and propose an extension of existing implementations in R. We illustrate the utility using two short psychological tests.
Keywords: Rasch model, R, graphics, item fit
Karl Bang Christensen,
Postboks 2099,
Øster Farimagsgade 5 opg. B,
1014 København K, 15 Øster Farimagsgade 5.
kach@sund.ku.dk
Detecting and Differentiating Extreme and Midpoint Response Styles in Rating-type scale using Tree-Based Item Response Models: Simulation Study and Empirical Evidence
Artur Pokropek, Lale Khorramdel & Matthias von Davier
Abstract
An extended the Tree-Based Item Response Models (IRTree) approach to detect response styles in Rating-type scale data and to differentiate between extreme response style (extreme response style) and midpoint response style is introduced and validated with a simulation study and using empirical data. The Tree-Based Item Response Models extension is based on the decomposition of rating data into binary pseudo items, which are examined using the multidimensional Item Response Theory modelling framework. Different scenarios, levels, and consistencies of extreme and midpoint response styles are simulated. The approach is further applied to selected scales of the PISA questionnaire. Results show that the approach is a useful and valid tool to detect and correct for response styles in Rating-type scales.
Keywords: large scale assessment, response styles, simulation study
Extending GMX: Conditional Likelihood Ratio Test and Extended Graphical Model Checks with psychotools
Rainer W. Alexandrowicz
Abstract
This article introduces an extension of GMX, which now also supports the conditional likelihood ratio test and graphical model checks for the psychotools package. The package is freely available at osf.io/2ryd8.
Keywords: Rasch models, graphical model check, conditional maximum likelihood, multi-group split, R-package, psychotools
Rainer W. Alexandrowicz,
University of Klagenfurt,
Institute of Psychology,
Methods Department,
Universitaetsstrasse 67–69,
9020 Klagenfurt, Austria
rainer.alexandrowicz@aau.at
The Need for Quantification
Theoretical considerations of a construct and evaluation of a scale for its measurement
Henrik Gast, Thomas Ostermann, Tugba Kapanci & Stefan J. Troche
Abstract
There are growing possibilities to quantify aspects of life, and an increasing number of individuals use smartphone applications for such quantifications. This leads us to assume that there is a need for quantification (NfQ). We define NfC as an individual's need to grasp numbers about their body, their experience and behavior, and self-related aspects of the individual's surroundings. In contrast to lifelogging or self-tracking, NfQ focuses on the motivational (and more general) level. In two studies with 375 and 216 participants, we developed and evaluated the 7-item NfQ scale to assess individual differences in NfQ. In both samples, the scale was unidimensional and highly reliable. In Study 1, the NfQ scale correlated with quantification-related behavior in different areas of life (documentation of sports activities, weight control, comparing prices, preference for feedback in form of grades/numbers), pointing to the breadth of NfQ-related behaviors. NfQ correlated positively with the need for cognitive closure and external control convictions but negatively with self-efficacy. This pattern of results suggests that NfQ-related behavior compensates for deficient control beliefs and is part of a reactive coping strategy to reduce tension in the face of ambivalence and hardly controllable situations. The NfQ scale might be a promising tool in education, sports training, health, medical and psychotherapeutic interventions when high NfQ can be used to increase commitment and motivation.
Keywords: need for quantification; scale; assessment; lifelogging; self-tracking
Prof. Dr. Stefan Troche,
Department of Psychology,
University of Bern,
Fabrikstr. 8,
CH-3012 Bern, Switzerland,
stefan.troche@unibe.ch
Breadth of Data Dispersion and Number of Variables as Sources of Measurement Effects on Factor Variances in Confirmatory Factor Analysis
Karl Schweizer, Stefan Troche and Tengfei Wang
Abstract
The present study investigates how simple measurement effects influence factor variances in CFA. The considered measurement effects refer to a) deviations from the expected dispersion of data and b) the number of manifest variables (e.g., items of a scale) loading on a factor while the underlying data structure is kept constant. In this investigation, the factor variance is conceptualized as the scaled variance parameter of the model-implied covariance matrix of the Maximum Likelihood approach. The results of model analyses and a simulation study revealed that the modification of the breadth of data dispersion and the number of manifest variables systematically influenced scaled factor variances despite the constancy of the latent structure. Furthermore, the results revealed that the effects on the factor variance could be eliminated by either estimating factor variances using standardized data or by standardizing estimated factor loadings before their conversion into factor variances.
Keywords: factor variance, dispersion, variable number, scaling, confirmatory factor analysis
Karl Schweizer,
Institute of Psychology,
Goethe University
Frankfurt, Germany
ORCID iD: 0000 0002 3143 2100
K.Schweizer@psych.uni-frankfurt.de
Psychological Test and Assessment Modeling
Volume 65 · 2023 · Issue 2
Pabst, 2023
ISSN 2190-0493 (Print)
ISSN 2190-0507 (Internet)