Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


osu1322364626.pdf (719.52 KB)

ETD Abstract Container

Abstract Header

The Performance of Local Dependence Indices with Psychological Data

Houts, Carrie Rena
http://rave.ohiolink.edu/etdc/view?acc_num=osu1322364626

Abstract Details

2011, Doctor of Philosophy, Ohio State University, Psychology.
The violation of the assumption of LI, known as local dependence (LD), when applying Item Response Theory (IRT) models has been shown to have a negative impact on all estimates from the model, specifically item-level, person-level, and test-level values. For this reason, numerous indices and statistics have been proposed to aid analysts in the detection of LD. A large-scale simulation study was conducted to evaluate the relative performance of selected LD measures, as measured by Type I error rate and power, as well as the parameter recovery for items in datasets exhibiting LD. Measures were assessed at differing levels of test length, number of respondents, type of LD present (i.e., none, surface, underlying), fitted IRT model, degree of LD exhibited by item pairs, and number of item pairs exhibiting LD. The levels of the independent variables were chosen to replicate conditions more typical of studies involving psychological assessment, as opposed to educational measurement. Parameter estimates when LD was present were negatively affected to such a degree as to support the recommendation to use LD indices as part of any preliminary analyses. Previous findings regarding differences between surface and underlying LD and the conformance of selected indices to their theoretical Null distributions were replicated. No LD index displayed the best performance across all conditions; two indices displayed superior performance in terms of both power and Type I error but are only available when the data is strictly dichotomous in nature. Two other LD measures are available across the two fitted IRT models examined and displayed adequate to good performance in most simulation conditions. The use of these indices in tandem is the final general recommendation for applied researchers. Areas of future investigation, including the observed power differences of LD indices when detecting surface versus underlying local dependence, are identified.
Michael Edwards, PhD (Advisor)
Thomas Nygren, PhD (Committee Member)
Robert Cudeck, PhD (Committee Member)
126 p.
Quantitative Psychology
IRT; local dependence; Monte Carlo; LD indices

Recommended Citations

Citations

  • Houts, C. R. (2011). The Performance of Local Dependence Indices with Psychological Data [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1322364626

    APA Style (7th edition)

  • Houts, Carrie. The Performance of Local Dependence Indices with Psychological Data. 2011. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1322364626.

    MLA Style (8th edition)

  • Houts, Carrie. "The Performance of Local Dependence Indices with Psychological Data." Doctoral dissertation, Ohio State University, 2011. http://rave.ohiolink.edu/etdc/view?acc_num=osu1322364626

    Chicago Manual of Style (17th edition)

osu1322364626
542
© 2011, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.