Login

Publications  •  Project Statistics

Glossary  •  Schools  •  Disciplines
People Search: 
   
Title/Abstract Search: 

Dissertation Information for Se-Kang Kim

NAME:
- Se-Kang Kim

DEGREE:
- Ph.D.

DISCIPLINE:
- Psychology

SCHOOL:
- University of Minnesota, Twin Cities (USA) (1999)

ADVISORS:
- Mark L. Davison

COMMITTEE MEMBERS:
- None

MPACT Status: Incomplete - Inspected

Title: A simulation comparing bootstrap and maximum likelihood estimates of scale-value standard errors in multidimensional scaling

Abstract: This study was conducted to find out the more appropriate way to estimate scale-value standard error (SE) in the analysis of profile data using the multidimensional scaling (MDS) technique, since no previous studies have been done. Many test batteries in education and psychology result in a profile of scores for each individual. However, the utility of score profiles has been largely neglected, and Davison (1996) called attention to that and developed a quantitative method for analyzing profile data utilizing MDS, which he called Profile Analysis via Multidimensional Scaling (PAMS). This study applied two well-known statistical techniques to PAMS for scale-value SE estimation; one is the bootstrap and the other is the maximum likelihood (ML) method. ML estimation is based on large sample theory (asymptotic theory), whereas the bootstrap is based on a resampling technique. Using the PAMS model, the study generated five hundred samples of the sample size 500. From each sample scale-values were estimated by both ALSCAL (nonmetric approach) and MULTISCALE (ML approach) and used to construct a scale-value sampling distribution for each method. Since MULTISCALE provides asymptotic SE, the bootstrap technique was applied to only the ALSCAL procedure to calculate SE. The standard deviation of the sampling distribution was considered "true" SE and used for iterion for comparison. The first 25 samples of the five hundred were analyzed both by the bootstrap and MULTISCALE . The study had three between-factors (i.e., dimensionality, number of tests, and error levels) and one within-factor (i.e., the bootstrap and the ML). Dependent variables were mean biases and one-sample t-tests, and they were computed by comparing 25 samples' SE estimates with the corresponding criteria. The results showed that over dimensionality, number of tests, and error levels 79% of bootstrap SE estimates were not significantly different from the criteria, whereas only 2% of ML estimates were not significantly different from the criteria. From the results, it is reasonable to conclude, at least in MDS of profile data, that bootstrap SE estimation is superior to ML SE estimation.

MPACT Scores for Se-Kang Kim

A = 0
C = 1
A+C = 1
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2010-11-14 14:39:12

Advisors and Advisees Graph