東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Dimensionality and invariance: Asse...

University of California, Los Angeles.

Linked to FindBook

Google Book

Amazon

博客來

Dimensionality and invariance: Assessing differential item functioning using bifactor multiple indicator multiple cause models.

Record Type:	Language materials, printed : Monograph/item
Title/Author:	Dimensionality and invariance: Assessing differential item functioning using bifactor multiple indicator multiple cause models./
Author:	Ainsworth, Andrew Thomas.
Description:	99 p.
Notes:	Adviser: Steven P. Reise.
Contained By:	Dissertation Abstracts International68-09B.
Subject:	Psychology, Psychometrics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoeng/servlet/advanced?query=3280990
ISBN:	9780549235286

Dimensionality and invariance: Assessing differential item functioning using bifactor multiple indicator multiple cause models.
Ainsworth, Andrew Thomas.

Dimensionality and invariance: Assessing differential item functioning using bifactor multiple indicator multiple cause models. - 99 p.

Adviser: Steven P. Reise.

Thesis (Ph.D.)--University of California, Los Angeles, 2007.

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.

ISBN: 9780549235286Subjects--Topical Terms:

1017742
Psychology, Psychometrics.

Dimensionality and invariance: Assessing differential item functioning using bifactor multiple indicator multiple cause models.
LDR:03185nam 2200265 a 45 001 857513
005 20100712
008 100712s2007 ||||||||||||||||| ||eng d
020 $a 9780549235286
035 $a (UMI)AAI3280990
035 $a AAI3280990
040 $a UMI $c UMI
100 1 $a Ainsworth, Andrew Thomas. $3 1024464
245 1 0 $a Dimensionality and invariance: Assessing differential item functioning using bifactor multiple indicator multiple cause models.
300 $a 99 p.
500 $a Adviser: Steven P. Reise.
500 $a Source: Dissertation Abstracts International, Volume: 68-09, Section: B, page: 6383.
502 $a Thesis (Ph.D.)--University of California, Los Angeles, 2007.
520 $a 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.
590 $a School code: 0031.
650 4 $a Psychology, Psychometrics. $3 1017742
690 $a 0632
710 2 $a University of California, Los Angeles. $3 626622
773 0 $t Dissertation Abstracts International $g 68-09B.
790 $a 0031
790 1 0 $a Reise, Steven P., $e advisor
791 $a Ph.D.
792 $a 2007
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoeng/servlet/advanced?query=3280990