2018
DOI: 10.1177/0013164418783530
|View full text |Cite
|
Sign up to set email alerts
|

Understanding the Model Size Effect on SEM Fit Indices

Abstract: This study investigated the effect the number of observed variables (p) has on three structural equation modeling indices: the comparative fit index (CFI), the Tucker-Lewis index (TLI), and the root mean square error of approximation (RMSEA). The behaviors of the population fit indices and their sample estimates were compared under various conditions created by manipulating the number of observed variables, the types of model misspecification, the sample size, and the magnitude of factor loadings. The results … Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

8
285
3
6

Year Published

2019
2019
2024
2024

Publication Types

Select...
8
1
1

Relationship

0
10

Authors

Journals

citations
Cited by 550 publications
(389 citation statements)
references
References 44 publications
(66 reference statements)
8
285
3
6
Order By: Relevance
“…Regardless, one strength of our analyses was that they were conducted at the item-level, which is arguably a stronger test of the structure of a given measure and can possibly better identify sources of model misspecification (as opposed to using subscale sum/mean scores as indicators of higher-order factors; Marsh, Lüdtke, Nagengast, Morin, & von Davier, 2013). One caveat with item-level factor analysis is that if the indicators in the model have less reliability (i.e., low factor loadings), fit index values may be biased, especially with many indicators (Shi, Lee, & Maydeu-Olivares, 2019) Disconstraint has evidence of relatively weaker coverage of maladaptive high Conscientiousness (e.g., Trull, Useda, Costa, & McCrae, 1995), as opposed to PiCD Anankastia having limited coverage. This is further supported by results in the original validation study of the PiCD, where PiCD Anankastia had relatively stronger negative relations with PID-5 Disinhibition and CAT-PD-SF Disconstraint (Oltmanns & Widiger, 2018).…”
Section: Discussionmentioning
confidence: 99%
“…Regardless, one strength of our analyses was that they were conducted at the item-level, which is arguably a stronger test of the structure of a given measure and can possibly better identify sources of model misspecification (as opposed to using subscale sum/mean scores as indicators of higher-order factors; Marsh, Lüdtke, Nagengast, Morin, & von Davier, 2013). One caveat with item-level factor analysis is that if the indicators in the model have less reliability (i.e., low factor loadings), fit index values may be biased, especially with many indicators (Shi, Lee, & Maydeu-Olivares, 2019) Disconstraint has evidence of relatively weaker coverage of maladaptive high Conscientiousness (e.g., Trull, Useda, Costa, & McCrae, 1995), as opposed to PiCD Anankastia having limited coverage. This is further supported by results in the original validation study of the PiCD, where PiCD Anankastia had relatively stronger negative relations with PID-5 Disinhibition and CAT-PD-SF Disconstraint (Oltmanns & Widiger, 2018).…”
Section: Discussionmentioning
confidence: 99%
“…We also randomly split the sample in two halves to perform an exploratory factor analysis (EFA) with one of the halves (N = 545) and a confirmatory factor analysis (CFA) with the other (N = 540). For the CFA, as Chi-square is influenced by sample size [59] and model size [60], and both are large in our study, so we relied more on the root mean square error of approximation (RMSEA), the comparative fit index (CFI) and the normed fit index (NFI) to check goodness of fit. Correlations with dark triad variables and a structural equation model (SEM) to predict judicial manipulation form dark traits were also calculated.…”
Section: Resultsmentioning
confidence: 99%
“…Although the dark factor has been described not only based on these variables, it is a robust construct and its predictive capacity is maintained even if important indicators are eliminated [23]. As chi-square is influenced by the sample size [47] and the size of the model [48], for CFA, we used other statistics to measure the goodness-of-fit, such as the comparative fit index (CFI), the standardized root mean square residual (SRMR), and the root mean square error of approximation (RMSEA). Zero-order correlations were calculated for all the variables in the model, which also served as an indicator of the concurrent validity of the revenge scale.…”
Section: Resultsmentioning
confidence: 99%