File(s) under permanent embargo
Examining Simultaneous Power in Structural Equation Modeling
thesis
posted on 2013-04-19, 00:00 authored by Robert Anthony PereraStatistical power is typically thought of univariately. However, the most common statistical models are multivariate, i.e., multiple regression and structural equation modeling. Researchers are likely interested in finding all parameters statistically significant that are non-null in the population. That is, it is desirable to commit no type II errors. The probability of committing no type II errors in an analysis can be termed simultaneous power. What is often ignored is that simultaneous can be substantially lower than the power for a single predictor, or more so if the probability of finding at least one significant result is examined. This expresses the need to explore ways to improve simultaneous power other than increasing sample size. This project examined four ways for simultaneously testing structural coefficients in structural equation modeling in the special case of two exogenous latent variables. Namely, these methods are the delta method, assymptotic confidence intervals based on the distribution of the products, the percentile bootstrap, and the bias-corrected and accelerated bootstrap.
History
Date Modified
2017-06-02Defense Date
2013-04-12Research Director(s)
Scott E. MaxwellCommittee Members
Ke-Hai Yuan Guangjian Zhang Lijuan WangDegree
- Doctor of Philosophy
Degree Level
- Doctoral Dissertation
Language
- English
Alternate Identifier
etd-04192013-135452Publisher
University of Notre DameProgram Name
- Psychology
Usage metrics
Categories
No categories selectedLicence
Exports
RefWorks
BibTeX
Ref. manager
Endnote
DataCite
NLM
DC