International Journal of Nonlinear Analysis and Applications

A proposed method for cleaning data from outlier values using the robust RFCH method in structural equation modeling

Document Type : Research Paper

Authors

1 College of Computer science and Mathematics, Mosul University, Mosul, Iraq

2 Statistics Department, College of Administration and Economics, University of Baghdad, Baghdad, Iraq

Abstract

The GLS and ML methods are the most common methods for estimating SEM but require multivariate normality. Therefore, methods robust to standard errors and quality of fit indexes to Chi-square have been proposed: MLR and they are considered superior to ML and GLS methods analyzing ordinal data. When we have a five-way Likert scale, the data is treated as continuous by calculating the covariance matrix as inputs for ML, GLS, and MLR. However, outliers are familiar because modeling requires a large sample size, either because of the input of data or answers more than is expressed within a particular category. Their presence affects even methods with robust corrections, where the accuracy of estimating parameters, standard errors, and fit indicators may be compromised the quality of fit indexes and inappropriate solutions, where a robust algorithm is proposed to clean the data from the outlier, as this proposed algorithm calculates the robust correlation matrix robust RFCH Reweighted Fast Consistent and High Breakdown, which consists of several steps and has been modified by taking the clean data before calculating the robust RFCH correlation matrix.  It was also suggested to make a comparison between the three methods before the treatment process with the presence of outlier values and note the extent of their impact on the methods and after using the robust RFCH method, and note the extent of improvement in estimations, standard errors and the overall quality of fit indexes for each of the Chi-square index, CFI, TLI, and RMSEA, SRMR and CRMR, with the robust corrections in the Chi-square index for each of the methods MLR. Through the simulation experiment, the researcher reached the power of the proposed method robust RFCH in improving the quality of parameter estimation, standard errors, and overall fit indexes quality.

Keywords

[1] P.M. Bentler and D.G. Bonett, Significance tests and goodness of fit in the analysis of covariance structures,
Psycho. Bull. 88 (1980) 588–606.
[2] P.M. Bentler, EQS 6 structural equations program manual, In Los Angeles: BMDP Statistic Software, 2006
[3] P.M. Bentler and P. Dudgeon, Covariance structure analysis: statistical practice, theory, and directions, Annual
Rev. Psycho. 47 (1996) 563–592.
[4] P.M. Bentler and K.H. Yuan, Structural equation modeling with small samples:Test statistics, Mult. Behav. Res.
34 (1999) 181–197.
[5] J. Biggs, Evaluation of the effect of outliers on the GFI quality adjustment index in structural equation models
and proposal of alternative quality indices, Commun. Stat. Simul. Comput. 46 (2013) 5–9.
[6] K.A. Bollen, Outliers and improper solutions: A confirmatory factor analysis example, Socio. Meth. Res. 15
(1987) 375–384.
[7] K.A. Bollen, Structural Equations with Latent Variables, John Wiley & Sons, 1989.
[8] P.E. Brosseau-Liard and V. Savalei, Adjusting incremental fit indices for nonnormality, Mult. Behav. Res. 49
(2014) 460–470.
[9] P.E. Brosseau-Liard, V. Savalei and L. Li, An investigation of the sample performance of two nonnormality
corrections for RMSEA, Mult. Behav. Res. 47 (2012) 904–930.
[10] M.W. Browne, Asymptotically distribution-free methods for the analysis of covariance structures, British J. Math.
Stat. Psycho. 37 (1984) 62–83.
[11] W.M. Browne, Generalized Least Squares Estimators in the Analysis of Covariance Structures, ETS Res. Bull.
Ser. 1973 (1) 1–36.[12] W.M. Browne and R. Cudeck, Alternative ways of assessing model fit, Socio. Meth. Res. 21 (1992) 230–258.
[13] B.M. Byrne, Structural Equation Modeling with Mplus: Basic Concepts, Applications and Programming, Routledge, 2013
[14] S. Cangur and I. Ercan, Comparison of model fit indices used in structural equation modeling under multivariate
normality, J. Modern Appl. Stat. Meth. 14 (2015) 152–167.
[15] C. Cheng and H. Wu, Confidence Intervals of Fit Indexes by Inverting a Bootstrap Test, Struct. Equ. Mod. 24
(2017) 870–880.
[16] M.A. Cirillo and L.P. Barroso, Effect of outliers on the GFI quality adjustment index in structural equation model
and proposal of alternative indices, Commun. Stat. Simul. Comput. 64 (2017) 1895–1905.
[17] J.E. Collier, Applied structural equation modeling using AMOS: Basic to advanced techniques, Routledge, 2020.
[18] A. Crisci, Estimation methods for the structural equation models: maximum likelihood, partial least squares and
generalized maximum entropy, J. Appl. Quant. Meth. 7 (2012) 7–10.
[19] S.J. Devlin, R. Gnanadesikan and J.R. Kettenring, Robust estimation of dispersion matrices and principal components, J. Amer. Stat. Assoc. 76 (1981) 354–362.
[20] M. Harwell, A strategy for using bias and RMSE as outcomes in Monte Carlo studies in statistics, J. Modern
Appl. Stat. Meth. 17 (2018).
[21] L. Hildreth, Residual Analysis for Structural Equation Modeling, Statistics, PhD., 2013.
[22] L.T. Hu and P.M. Bentler, Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria
versus new alternatives, Struct. Equ. Model. 6 (1999) 1–55.
[23] S.R. HutChinson and A. Olmos, Behavior of descriptive fit indexes in confirmatory factor analysis using ordered
categorical data, Struct. Equ. Model. 5 (1998) 344–364.
[24] S. Jalal and P.M. Bentler, Using Monte Carlo normal distributions to evaluate structural models with nonnormal
data, Struct. Equ. Model. 25 (2018) 541–557.
[25] R.B. Kline, Principles and practices of structural equation modelling, In Methodology in the social sciences, (4th
ed.). Guilford publications, 2016.
[26] S.Y. Lee, Structural Equation Modeling A Bayesian Approach, John Wiley & Sons, 2007.
[27] C.H. Li, The performance of MLR, USLMV, and WLSMV estimation in structural regression models with ordinal
variables, J. Chem. Inf. Model. 53 (2014) 1689–1699.
[28] R. J. Magnus and H. Neudecker, Matrix Differential Calculus with Applications in Statistics and Econometrics,
John Wiley & Sons, 2019.
[29] A. Maydeu-Olivares, D. Shi and Y. Rosseel, Assessing fit in structural equation models: A Monte-Carlo evaluation
of RMSEA versus SRMR confidence intervals and tests of close fit, Struc. Equ. Model. 25 (2018) 389–402.
[30] H. Midi, H.T. Hendi, J. Arasan and H. Uraibi,Fast and robust diagnostic technique for the detection of high
leverage points, Pertanika J. Sci. Tech. 28 (2020) 1203–1220.
[31] H. Ogasawara, Standard errors of fit indices using residuals in structural equation modeling, Psychometrika 66
(2001) 421–436.
[32] D.J. Olive and D. Hawkins, Robust multivariate location and dispersion, Unpublished Manuscript Available From
(Http://Www. Math. Siu. Edu/Olive/Pphbmld.Pdf), (2010) 1–30.
[33] D.J. Olive, Robust Multivariate Analysis, Springer, 2017.
[34] D.J. Olive and D.M. Hawkins, High breakdown multivariate estimators, (2008) 1–29.
[35] J. Palomo, D.B. Dunson and K. Bollen, Bayesian Structural Equation Modeling, In Handbook of Latent Variable
and Related Models, 2007.
[36] M. Rhemtulla, P.E. Brosseau-Liard and V. Savalei, When can categorical variables be treated as continuous? A
comparison of robust continuous and categorical SEM estimation methods under suboptimal conditions, Psycho.
Meth. 17 (2012) 354–373.
[37] E.E. Rigdon, CFI versus RMSEA: A comparison of two fit indexes for structural equation modeling, Struct. Equ.
Model. 3 (1996) 369–379.
[38] P.J. Rousseeuw and K. VanDriessen, A fast algorithm for the minimum covariance determinant estimator, Technom. 41 (1999) 212–223.
[39] V. Savalei, Expected versus observed information in SEM with incomplete normal and nonnormal data, Psycho.
Meth. 15 (2010) 352–367.
[40] V. Savalei, Understanding robust corrections in structural equation modeling, Struct. Equ. Model. 21 (2014) 149–
160.
[41] K. Schermelleh-Engel, H. Moosbrugger and H. M¨uller, Evaluating the fit of structural equation models: Tests of
significance and descriptive goodness-of-fit measures, MPR-Online, 8 (2003) 23–74.
[42] R.E. Schumacker and R. G. Lomax,A Beginner’s Guide to Structural Equation Modeling, New York, NY: Taylorand Francis Group, LLC, 2010.
[43] J.D. Schweig, Multilevel factor analysis and student ratings of instructional practice, UCLA. 2014.
[44] D. Shi, A. Maydeu-Olivares and Y. Rosseel, Assessing Fit in Ordinal Factor Analysis Models: SRMR vs. RMSEA,
Struct. Equ. Model. 27 (2020) 1–15.
[45] B.G. Tabachnick and L.S. Fidell, Using Multivariate Statistics, Pearson Boston, 2014
[46] N.H. Timm, Applied Multivariate Analysis, Springer,New York Berlin Heidelberg, 2002
[47] X. Tong, Evaluation of the Robustness of Modified Covariance Structure Test Statistics, University of California,
Los Angeles, 2012
[48] H. S. Uraibi and H. Midi, On robust bivariate and multivariate correlation coefficient, Econ. Comput. Econ.
Cyber. Stud. Res. 53 (2019) 221–239.
[49] H.S. Uraibi, H. Midi and S. Rana, Robust multivariate least angle regression, Sci. Asia 43 (2017) 56–60.
[50] C.D. Vale and V.A. Maurelli, Simulating multivariate nonnormal distributions, Psychometrika 48 (1983) 465–471.
[51] J. Wang and X. Wang,Structural equation modeling: Applications using Mplus, John Wiley & Sons, 2020.
[52] Y. Xia and Y. Yang, RMSEA, CFI, and TLI in structural equation modeling with ordered categorical data: The
story they tell depends on the estimation methods, Behav. Res. Meth. 51 (2019) 409–428.
[53] F. Yang-Wallentin, K.G. J¨oreskog and H. Luo, Confirmatory factor analysis of ordinal variables with misspecified
models, Struct. Equ. Model. 17 (2010) 392–423.
[54] K.H. Yuan and P. Bentler, Normal theory based test statistics in structural equation modelling, British J. Math.
Stat. Psycho. 51 (1998) 289–309.
[55] K. H. Yuan and P. Bentler, Effect of outliers on estimators and tests in covariance structure analysis, British J.
Math. Stat. Psycho. 54 (2001) 161–175.
[56] K. H. Yuan, P. M. Bentler and W. Zhang, The effect of skewness and kurtosis on mean and covariance structure
analysis: The univariate case and its multivariate implication, Soc. Meth. Res. 34 (2005) 240–258.
[57] K. H. Yuan, W. Chan and P.M. Bentler, Robust transformation with applications to structural equation modelling,
British J. Math. Stat. Psycho. 53 (2000) 31–50.
[58] K. H. Yuan and K. Hayashi, Standard errors in covariance structure models: Asymptotics versus bootstrap, British
J. Math. Stat. Psycho. 59 (2006) 397–417.
[59] J. Zhang, Applications of a robust dispersion estimator, Doctoral dissertation, Southern Illinois University Carbondale, 2011.
[60] J. Zhang, D. J. Olive and P. Ye,Robust covariance matrix estimation with canonical correlation analysis, Int. J.
Stat. Prob. 1 (2012) 119–136.
International Journal of Nonlinear Analysis and Applications
Volume 12, Issue 2
November 2021
Pages 2269-2293
  • Receive Date: 17 March 2021
  • Revise Date: 16 August 2021
  • Accept Date: 02 September 2021