|
Psychometric modeling of response process heterogeneity with mixture item response tree models
Alagöz, Ömer Emre Can
|
URN:
|
urn:nbn:de:bsz:180-madoc-721626
|
|
Dokumenttyp:
|
Dissertation
|
|
Erscheinungsjahr:
|
2026
|
|
Ort der Veröffentlichung:
|
Mannheim
|
|
Hochschule:
|
Universität Mannheim
|
|
Gutachter:
|
Meiser, Thorsten
|
|
Datum der mündl. Prüfung:
|
2026
|
|
Sprache der Veröffentlichung:
|
Englisch
|
|
Einrichtung:
|
Fakultät für Sozialwissenschaften > Psychologische Methodenlehre u. Diagnostik (Meiser 2009-)
|
|
Lizenz:
|
Creative Commons Namensnennung 4.0 International (CC BY 4.0)
|
|
Fachgebiet:
|
150 Psychologie
|
|
Freie Schlagwörter (Englisch):
|
mixture models , response styles , IRTree models , psychometrics
|
|
Abstract:
|
The use of rating scale items in self-report data collection is an economic and convenient way to collect large amounts of data. However, the validity of the responses are threatened by heuristic processes such as response styles (RS). RS are systematic tendencies to select specific response categories regardless of the item content, helping respondents to reduce the effort for responding, and may bias trait scores, factor structures, correlations and rank ordering of persons. Although there are several psychometric models that aim to disentangle the RS effects, they bear a key limitation: the assumption of homogeneous response processes across individuals. That is, all respondents are assumed to use RS while responding, and their responses are affected by RS to the same extent. In case this assumption is not met, the previous psychometric models fall short as they do not account for such heterogeneity and fail correcting for RS styles, thus still result in biased estimates.
In this dissertation, using the state-of-art methodology for accounting response styles, I first develop a confirmatory mixture model that tests the hypothesis of homogeneous response processes (i.e., Article I). I found that respondents do not only differ whether they use or do not use RS but also employ different combinations of them. These findings highlight that assumption of homogeneity is not always tangible and the heterogeneity of response processes may be more complicated than it first appears.
Following upon, I develop a more flexible mixture framework for modeling process heterogeneity (i.e., Article II). The so-called MixTree model allows any number of finite mixture components where the weighting of trait and RS can differ between latent sub-populations. Furthermore, through a stepwise-maximum-likelihood estimation, the MixTree allows prediction of latent class memberships to explore potential sources of the
observed heterogeneity.
Simultaneously, a new methodology for capturing the response process heterogeneity arose in the literature, score-based partitioning (SBP) algorithm. Using machine-learning techniques, SBP finds subgroups differing in response processes using covariate information.
In Article III, I compare the mixture and SBP approaches conceptually and technically. This work is the first one that compares the methods for a psychometric model
that investigates response processes to our knowledge. Findings show a clear picture of which of the two methods is preferable under certain circumstances, serving as a guide for applied researchers and foundation for further methodological research.
|
 | Dieser Eintrag ist Teil der Universitätsbibliographie. |
 | Das Dokument wird vom Publikationsserver der Universitätsbibliothek Mannheim bereitgestellt. |
 Suche Autoren in
Sie haben einen Fehler gefunden? Teilen Sie uns Ihren Korrekturwunsch bitte hier mit: E-Mail
Actions (login required)
 |
Eintrag anzeigen |
|
|
