- Title
- Index optimisation for structural equation models (SEM)
- Creator
- Stindt, Carmen
- Subject
- Structural equation modelling
- Date Issued
- 2018
- Date
- 2018
- Type
- Thesis
- Type
- Masters
- Type
- MCom
- Identifier
- http://hdl.handle.net/10948/17919
- Identifier
- vital:28518
- Description
- Structural equation modelling (SEM), a statistical technique used extensively in quantitative marketing research and other domains, is an analytical approach used to model latent (unobservable) variables. Unlike distribution fitting where simple chi-squared goodness-of-fit assessment yields satisfactory results, model fit in SEM is more difficult. Descriptive goodness-of-fit indices have been developed over the past 50 years to assist in the assessment of model fit. The traditional assessment method requires reporting multiple indices, all of which should reflect an adequate model fit in order for the overall model fit to be deemed good. The choice of indices to report are left to the researcher’s discretion, leading to the indices used to differ considerably. The combination of using the traditional assessment method and differing indices often lead to conflicting results. This study proposes a composite index, combining frequently used indicators in an attempt to obtain a single index method for assessing model fit in SEM that performs better when compared to the traditional assessment method. Composite indices have been used in other domains as an improved method of assessing performance (Barr and Kantor, 2004). The composite index proposed is evaluated using a Monte Carlo simulation study under different experimental conditions. The experimental conditions investigated are sample size, estimation method and model misspecification. These experimental conditions are chosen to investigate as each has been shown to affect the traditional indices performances. The ideal fit indices should be able to detect model misspecification while being insensitive to sample size and estimation methods. This is not always the case with the traditional indices. The composite index proposed is shown to outperform the traditional assessment method under many of the experimental condition combinations. This provides evidence that composite indices may be a more beneficial method of assessing model fit in SEM.
- Format
- x, 112 leaves
- Format
- Publisher
- Nelson Mandela University
- Publisher
- Faculty of Science
- Language
- English
- Rights
- Nelson Mandela University
- Hits: 758
- Visitors: 829
- Downloads: 118
Collections
NMU School of Computer Science, Mathematics, Physics and Statistics
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | SOURCE1 | Adobe Acrobat PDF | 2 MB | Adobe Acrobat PDF | View Details Download |