Naslov
Latent class analysis of the revised Short Test of Music Preferences on a Croatian sample

Autori
Prot, Sara ; Banožić, Adrijana ; Bosnar, Ksenija ; Prot, Franjo

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
20th Ramiro and Zoran Bujas Days - Book of Abstracts / Dinka-Čorkalo Biruški, Dinka ; Vukasović, Tena - Zagreb : Odsjek za psihologiju Filozofskog fakulteta Sveučilišta u Zagrebu, 2011, 37-37

ISBN
978-953-175-391-3

Skup
20th Ramiro and Zoran Bujas Days

Mjesto i datum
Zagreb, Hrvatska, 07.04.2011. - 09.04.2011

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Domaća recenzija

Ključne riječi
The Short Test of Music Preferences ; STOMP ; Croatian Translation ; Hopfield Neural networks Factor analysis ; Component analysis ; Promax rotation ; Clusterin ; K-means clustering ; Ward hierarchical clustering ; Morphotax algorithm of polar taxa ; Triatlon clustering by neural networks ; Classification evaluation ; Discriminant analysis

Sažetak
The Short Test of Music Preferences (STOMP ; Rentfrow & Gosling, 2003) has been developed to measure liking of different music genres. In previous studies on American samples it has been shown to measure four broad music-preference dimensions: Intense & Rebellious ; Reflective & Complex ; Upbeat & Conventional ; and Energetic & Rhythmic. Croatian translation of revised version of STOMP was used on a sample of 1005 university students to explore possible intercultural differences in the structure of music preferences. To obtain the latent structure in the space of STOMP items, component analysis with promax transformation was done, with PB factor retention criterion (Štalec & Momirović, 1971). Four extracted factors clearly corresponded to the structure suggested by Rentfrow & Gosling (2006 ; 2003). Latent class analysis was conducted by four taxonomic algorithms representing agglomerative, hierarchical and polar taxonomic approach ; i.e. K- means procedure (MacQueen, 1967), Ward method of hierarchical clustering (Ward, 1963), MORFOTAX algorithm for detection of polar taxa (Szirovicza & al., 1978), and TRIATLON algorithm for detection of clusters by neural networks (Momirović, 2003). The number of clusters was fixed corresponding to PB criterion. Efficiency of classification was evaluated by a series of discriminant analyses in manifest and latent space, showing statistically significant differences between groups defined by all four algorithms, after Bonferroni correction of probabilities was applied. Interpretation of taxonomic solutions was based on comparisons of group mean vectors of initial variables, as well as of 4 factors. Different algorithms produced different, but interpretable solutions. MORFOTAX algorithm produced solution where three of four dimensions matched those obtained in the factor solution.

Izvorni jezik
Engleski

Znanstvena područja
Psihologija

Napomena
Contribution is presented at: 5th psychometric symposium in honour of Professor Alija Kulenović Convenor: Vesna Buško Department of Psychology, Faculty of Humanities and Social Sciences, University of Zagreb, Croatia It is a great honor to announce the fifth international symposium dedicated to the memory of Professor Alija Kulenović. Organizing the symposium served as still another opportunity to remind ourselves to the work and authority of this vigorous teacher, creative scientist and outstandingly generous personality. The program of the present psychometric event covers diverse conceptual and methodological issues treated within the framework of classical and/or modern formulations of psychometric theory, or with reference to substantive phenomena, related operationalizations, and implemented validation procedures. Albeit the symposium is not devoted to a particular thematic area, I am pleased to note that the studies to be presented deal with theoretically and practically important questions related to the measurement of psychological attributes and possible methodological solutions. Typical construct validation problems are described in several studies using different analytic procedures, such as latent class analysis (S. Prot et al.), exploratory and confirmatory factor analyses (G. Milas et al., and A. Babić), or multigroup CFA (V. Takšić et al.). Two presentations include applications of classical test theory models and its generalizations in resolving particular estimation problems (B. Rebernjak and D. Urch), or in testing substantive hypotheses from specific applied areas (A. Mujagić). Also, demonstrations of use of new analytical solutions like local structural equation models (Hildebrandt et al.), and pairwise comparison method in the measurement of latent variables (A. Thomas et al.) are offered and exemplified with real and simulated data. Finally, the issues of metric quality of item and/ or composite test scores as a function of a mode of administration of instruments (L. Komidar and G. Sočan), item format (D. Ljubotina), or scoring method used (V. Buško) will also be presented. We hope that the contributions will serve to rouse interesting psychometric discussions in the audience and turn the symposium into an inspiring scientific enterprise.

POVEZANOST RADA