東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Consequences of violating the parame...

Coffman, Donna L.

Linked to FindBook

Google Book

Amazon

博客來

Consequences of violating the parameter drift assumption in covariance structure models.

Record Type:	Language materials, printed : Monograph/item
Title/Author:	Consequences of violating the parameter drift assumption in covariance structure models./
Author:	Coffman, Donna L.
Description:	144 p.
Notes:	Adviser: Robert C. MacCallum.
Contained By:	Dissertation Abstracts International67-01B.
Subject:	Psychology, Psychometrics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3200749
ISBN:	9780542463648

Consequences of violating the parameter drift assumption in covariance structure models.
Coffman, Donna L.

Consequences of violating the parameter drift assumption in covariance structure models. - 144 p.

Adviser: Robert C. MacCallum.

Thesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2005.

This dissertation involves an examination of the parameter drift assumption underlying the testing and evaluation of model fit in covariance structure models (CSMs). The assumption states that as sample size increases the degree to which the model is incorrect in the population (model error) decreases. This assumption is never valid in practice because the degree to which the model is wrong in the population will not change by increasing the sample size. Nearly all previous research on model misspecification has generated model error of a parametric nature by deleting or adding paths or latent variables. In contrast, I used a technique to generate a covariance matrix with a specified discrepancy function value in the population. This model misspecification does not have any particular parametric structure. Using simulation methods, I varied the degree of misfit, sample size; and model structure and examined how closely the empirical distribution of the test statistic followed a noncentral chi-square distribution, the degree of bias of the root mean square error of approximation (RMSEA), the relative noncentrality index (RNI), and the comparative fit index (CFI) point estimates, the coverage proportions of the RMSEA confidence intervals (CIs), the Type I error and power of the test of close fit, and power of the test of exact fit under violations of the assumption. Across the conditions studied, the mean and variance of the empirical distribution did not differ significantly from the mean and variance of the theoretical noncentral chi-square distribution. The CFI and RNI were unbiased in nearly every condition. The RMSEA point estimates underestimated the population RMSEA in most conditions. However, the coverage proportions for the RMSEA CIs maintained the coverage rate well. The Type I error rate for the test of close fit maintained the nominal level. The power for the test of close fit matched the theoretical power well. Thus, the results indicate robustness to violations of the parameter drift assumption for the conditions examined here. These results were unexpected based on previous research. It is speculated that the method for introducing model error is the reason for the unexpected results.

ISBN: 9780542463648Subjects--Topical Terms:

1017742
Psychology, Psychometrics.

Consequences of violating the parameter drift assumption in covariance structure models.
LDR:03162nam 2200277 a 45 001 969891
005 20110920
008 110921s2005 eng d
020 $a 9780542463648
035 $a (UnM)AAI3200749
035 $a AAI3200749
040 $a UnM $c UnM
100 1 $a Coffman, Donna L. $3 1293947
245 1 0 $a Consequences of violating the parameter drift assumption in covariance structure models.
300 $a 144 p.
500 $a Adviser: Robert C. MacCallum.
500 $a Source: Dissertation Abstracts International, Volume: 67-01, Section: B, page: 0594.
502 $a Thesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2005.
520 $a This dissertation involves an examination of the parameter drift assumption underlying the testing and evaluation of model fit in covariance structure models (CSMs). The assumption states that as sample size increases the degree to which the model is incorrect in the population (model error) decreases. This assumption is never valid in practice because the degree to which the model is wrong in the population will not change by increasing the sample size. Nearly all previous research on model misspecification has generated model error of a parametric nature by deleting or adding paths or latent variables. In contrast, I used a technique to generate a covariance matrix with a specified discrepancy function value in the population. This model misspecification does not have any particular parametric structure. Using simulation methods, I varied the degree of misfit, sample size; and model structure and examined how closely the empirical distribution of the test statistic followed a noncentral chi-square distribution, the degree of bias of the root mean square error of approximation (RMSEA), the relative noncentrality index (RNI), and the comparative fit index (CFI) point estimates, the coverage proportions of the RMSEA confidence intervals (CIs), the Type I error and power of the test of close fit, and power of the test of exact fit under violations of the assumption. Across the conditions studied, the mean and variance of the empirical distribution did not differ significantly from the mean and variance of the theoretical noncentral chi-square distribution. The CFI and RNI were unbiased in nearly every condition. The RMSEA point estimates underestimated the population RMSEA in most conditions. However, the coverage proportions for the RMSEA CIs maintained the coverage rate well. The Type I error rate for the test of close fit maintained the nominal level. The power for the test of close fit matched the theoretical power well. Thus, the results indicate robustness to violations of the parameter drift assumption for the conditions examined here. These results were unexpected based on previous research. It is speculated that the method for introducing model error is the reason for the unexpected results.
590 $a School code: 0153.
650 4 $a Psychology, Psychometrics. $3 1017742
650 4 $a Statistics. $3 517247
690 $a 0463
690 $a 0632
710 2 0 $a The University of North Carolina at Chapel Hill. $3 1017449
773 0 $t Dissertation Abstracts International $g 67-01B.
790 $a 0153
790 1 0 $a MacCallum, Robert C., $e advisor
791 $a Ph.D.
792 $a 2005
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3200749