Preface

In Memoriam of Professor Wang Wen Chung

Noncognitive variables refer to psychological constructs other than intelligence, achievement, or knowledge. Examples of such variables are, among others, attitude, personality, interests, motivation, beliefs, and emotions. Given their importance as predictors or as outcomes in their own right, these construct require a “better measurement” (Lipnevich, Preckel, & Roberts, 2016). In recognizing this need, this special issue of Applied Psychological Measurement brings together contributors’ expertise toward advances in the measurement of such variables.

Andrich and Luo demonstrate how two distinct set of instructions (preference vs. Judgment) elicit two different comparison processes that require different psychometric models for their analysis and interpretation. The authors showed that the hyperbolic cosine model (HCM) is the most appropriate for preference data. The implications of the authors’ Law of Comparative Preference for the measurement of attitudes, beliefs, and the like are straightforward: stating the right instruction, writing appropriate items, and modeling the responses accordingly through HCM.

Buchholz and Hartig propose a new item response theory (IRT)–based approach for studying the measurement invariance of attitudes needed for cross-country comparison in large-scale assessment programs. Using the cumulative generalized partial credit model (GPCM), data were simulated and their invariance explored via the root mean square deviance (RMSD) item-fit statistic. Results are promising for the study of noninvariance in cross-country comparisons.

Lee, Joo, Stark, and Chernyshenko develop a Markov Chain Monte Carlo (MCMC) method for estimating the parameters of the generalized graded unfolding–RANK model generated from the authors’ multiunidimensional pairwise preference (MUPP) model. The authors explore the parameters in a forced choice personality measure. Relevant implications for practitioners constructing forced choice measures are presented.

Liu and Wang address the important issue of response set (e.g., acquiescence), so common in noncognitive measures (e.g., attitudes). The authors propose the general unfolding model for response styles (GUMRS), which takes into account response set for unfolding data. The usefulness and robustness of the GUMRS for accommodating response styles is demonstrated with empirical and simulated data along with suggestions for fitting GUMRS to real unfolding data.

McGrane addresses a fundamental controversy as for the nature of attitudes. Bipolar models or bivariate ones? By using the general HCM for polytomous responses with empirical data, the author is able to show that a bivariate understanding of attitudes is flawed due to the use of inappropriate measurement models. This is especially relevant for the measurement of ambivalent attitudes. The suggestion of the author of carefully defining a construct and then using the appropriate measurement model cannot be more pertinent to the goals of this special issue (cf. Andrich & Luo).

These articles provide interesting psychometric models and estimation procedures to be employed by researches from a variety of applied fields working with noncognitive variables. One important takeaway of this special issue is the recognition of the relevance of (a) theoretically defining the nature of the construct, (b) considering the type of comparison process elicited in the subjects responding to an item, and (c) choosing the appropriate psychometric model for the appropriate generated data.