Dimensionality and Invariance: Assessing Differential Item Functioning Using Bifactor Multiple Indicator Multiple Cause Models
- ISBN-10
- 0549235280
- ISBN-13
- 9780549235286
- Category
- Factor analysis
- Pages
- 198
- Language
- English
- Published
- 2007
- Author
- Andrew Thomas Ainsworth
Description
The field of psychology is increasingly more ambitious in its attempts to study psychological phenomenon in diverse populations and demonstrate that measures are assessing a trait or construct equally across multiple groups. Correspondingly, there has been an increase in studies addressing differential item functioning (DIF) which is defined as item level group differences after controlling for the underlying trait distribution. For example, DIF occurs when 2 people with the same trait level (e.g. depression) but in different groups have different mean levels or endorsement probabilities on a specific item or items that measures a trait. A problem is that researchers often identify DIF and assume that the items can be modeled with a single dimension when even highly unidimensional scales can have "nuisance" secondary factors that can contaminate DIF detection. If data has a multidimensional (e.g. bifactor) structure but is assessed unidimensionally the identification of DIF becomes problematic at best and leads to a number of potentially spurious invariance violations. The current analysis investigated the influence of secondary factors on the assessment of DIF by simulating data with a bifactor structure and then investigating DIF using both a unidimensional multiple indicator multiple cause (MIMIC) model and a bifactor MIMIC model. Dichotomous item responses were simulated while manipulating the size of the group mean differences (3 levels), the size of the general factor loadings (3 levels), the size of the specific factor loadings (3 levels) and the combinations of mean differences on the general and specific factors (4 levels) for a total of 108 conditions. An additional 9 control conditions were simulated where no group differences were specified. Results indicate that an unmodeled subfactor structure (e.g. bifactor) is a viable cause of a high rate of spurious DIF detection and the proper modeling of the underlying factor structure can substantially reduce if not eliminate the occurrence of spurious DIF detection and lead to lower type 1 error rates. The current study shows that it is critically important to investigate and identify the underlying dimensionality of an item set before proceeding with DIF analyses.
Get the book
Similar books
-
Scale Invariance and Beyond: Les Houches Workshop, March 10–14, 1997
By B. Dubrulle, F. Graner, D. SornetteThis exhibits very clearly the respective roles of invariance and renormalization in such a Rescaling Symmetry. The mathematical treatment parallels very closely the arguments from Dimensionality that were recalled in the introduction, ...
-
Theory of Technical Change and Economic Invariance: Application of Lie Groups
By Ryuzo Satothat p} in (1) or (1") is a uniform magnification group with # = p and that p satisfies the “invariance” condition U"p – p = 0. (3-iii) (A4) Dimensional Invariance Axiom." A dimensional change in the unit of the currency does not change ...
-
Group Invariance in Statistical Inference
By Narayan C. GiriThis , however , does not reduce the dimension of the parametric space . By introducing the group invariance principle and restricting attention to invariant decision rules a reduction to the dimension of the parametric space is ...
-
Statistical Approaches to Measurement Invariance
By Roger E. MillsapMH.procedure.makes. no.dimensionality.assumptions.does.not.mean.that.the.procedure. ... Evans.simulated. item.scores.using.a.two-dimensional.compensatory. ... The. item. response. functions. were. invariant. over. groups.
-
Cosmological Crossroads: An Advanced Course in Mathematical, Physical and String Cosmology
By Spiros Cotsakis, Eleftherios Papantonopoulosdimensions, it suffices to notice that the conservation of the stress tensor Tαβ, which is a consequence of two-dimensional reparametrization invariance, in light-cone coordinates reads: ∂−T+++∂+T−+ = 0. In view of T−+ = T+− = 0, ...
-
Yang-baxter Equations, Conformal Invariance And Integrability In Statistical Mechanics And Field Theory - Proceedings Of A Conference
By Barber Michael N, Pearce Paul AConformal invariance in one-dimension implies the correlation functions must be constant. It is demonstrated by an exact solution that all the correlations between like species in a two-component lattice gas with the logarithmic ...
-
Foundations of Measurement: Representation, axiomatization, and invariance
By Patrick Suppes, David H. Krantz, Robert Duncan LuceEach subset of R in which p, q, r are fixed and 'r varies throughout Re is called a dimension. ... proved in Section 10.3, asserts that every dimensionally invariant relation on physical quantities x1, . . . , xk can be written as ...
-
Scale Invariance: Self-Similarity of the Physical World
By Richard N. HenriksenThese are often referred to as 'dimensions', which we shall write with a small 'd'. ... In principle, using Kilo-Dollars or Dollars to describe the price of an item leaves the price 'invariant', although the psychological reaction to ...
-
Marketing of Tourism Experiences
By Noel Scott, Eric Laws, Philipp BoksbergerThey consider visiting a major attraction (peak experiences), eating a gelato, and using transportation (supporting experiences) as an experience. They acknowledge the existence of staged experiences. Similarly, from an examination of ...
-
Conformal Invariance and Applications to Statistical Mechanics
By C Itzykson, H Saleur, J-B ZuberJK, 75.40,--s Conformal invariance powerfully constrains the critical behavior of two-dimensional classical (and onedimensional quantum) systems.” Critical theories are parametrized by the conformal anomaly c, which is the central ...