Please use this identifier to cite or link to this item: https://ah.lib.nccu.edu.tw/handle/140.119/118984
題名: CB-SEM和PLS-SEM在估計交互作用效果之比較
The comparison of estimation accuracy in interaction effect between CB-SEM and PLS-SEM
作者: 李昭鋆
Lee, Chao-Yun
貢獻者: 余民寧
Yu, Ming-Ning
李昭鋆
Lee, Chao-Yun
關鍵詞: CB-SEM
PLS-SEM
交互作用
CB-SEM
PLS-SEM
Interaction effect
日期: 2018
上傳時間: 27-Jul-2018
摘要: 本研究研究目的主要在於瞭解CB-SEM和PLS-SEM之指標乘積法、正交法、二階段法、無限制法在結構方程式模型之交互作用中,對迴歸係數、因素負荷量、解釋量、因素分數估計結果之良窳;此外,並瞭解指標數目、人數、資料類型對估計之影響。故本研究之實驗情況,共分三種指標數目、九種人數、二種資料類型,合計五十四種類型,並在每一類型模擬五百次。研究結果顯示,以整體論,在大部份的情況下,CB-SEM在迴歸係數、解釋量、因素負荷量表現較佳,而PLS-SEM在估計因素分數上較佳。而精確來說,若研究目的乃欲精確估計因數分數,則三百人、四題以上,建議採取PLS-SEM二階段法;若研究目是在精確估計迴歸係數、解釋量,若人數在四百人以上,建議採用CB-SEM無限制法。另外,本研究亦發現在大部份的情況下,指標數目愈多,資料型態為連續型態者,其估計效果愈佳。
The purpose of this study is to find out the results of estimation about interaction effects. The estimations come from the product indicator, two stage, orthogonalizing, and unconstrained approach which are estimated by CB-SEM and PLS-SEM separately. In the research, regression coefficient, factor loading, factor score, and r square are calculated by eight kinds of methods. Besides, night kinds of sample sizes, three kinds of number of indicators, and two kinds of data types are also studied to realize how they influence the estimations. Therefore, there are fifty-four situations. The results show that the estimation of regression coffoeicient, r square and factor loading are excellent by CB-SEM, but the estimation of factor score is better by PLS-SEM under most situations. If the object is to estimate the factor score, two satge approach of PLS-SEM is suggested based on the condition of sample size larger than 300 and 4 indicators. However, if the aim is to estimate the regression coffoeicient , r square, or factor loading, the unconstrained method of CB-SEM is the best choice when sample size is larger than 400. In addition, the continuous data type and more numbers of indicators are good for estimation under most conditions.
參考文獻: 朱經明(2007)。教育統計學 。臺北市:五南。\r\n余民寧(2006)。潛在變項模式:SIMPLIS的應用。臺北市: 高等教育。\r\n邱皓政(2003)。結構方程模式 : LISREL的理論、技術與應用。臺北\r\n市: 雙葉書廊。\r\n陳順宇 (2007)。結構方程模式 : Amos操作。臺北市: 心理總經銷.\r\n黃文璋 (2003)。數理統計。臺北市: 華泰.\r\n黃芳銘 (2010)。結構方程模式理論與應用。臺北市: 五南.\r\n蕭文龍 (2015)。統計分析入門與應用-SPSS中文版+PLS-\r\nSEM(SmartPLS)。臺北市:碁峰。\r\nAiken, L. S., & West, S. G. (1991). Multiple regression:\r\nTesting and interpreting interactions. Newbury Park,\r\nCA: Sage.\r\nBartholomew, D. J., & Knott, M. (1999). Latent variable\r\nmodels and factor analysis. New York: Oxford University\r\nPress.\r\nBollen, K. A. (1989). Structural equations with latent\r\nvariables. New York: Wiley.\r\nBollen, K. A., & Paxton, P. (1998). Two-stage least\r\nsquares estimation of interaction effects. In R. E.\r\nSchumacker & G. A. Marcoulides (Eds.), Interaction and\r\nnonlinear effects in structural equation modeling.\r\nMahwah, NJ: Lawrence Erlbaum Associates.\r\nBrandt, H., Kelava, A., & Klein, A. (2014). A simulation\r\nstudy comparing recent approaches for the estimation of\r\nnonlinear effects in sem under the condition of\r\nnonnormality. Structural Equation Modeling: A\r\nMultidisciplinary Journal, 21(2), 181-195.\r\nChen, C. (2016). The role of resilience and coping styles\r\nin subjective well-being among chinese university\r\nstudents. Asia-Pacific Education Researcher, 25(3),\r\n377-387.\r\nChin, W. W., Marcolin, B. L., & Newsted, P. R. (2003). A\r\npartial least squares latent variable modeling approach\r\nfor measuring interaction effects: Results from a Monte\r\nCarlo simulation study and an electronic-mail\r\nemotion/adoption study. Information Systems Research,\r\n14(2), 189-217.\r\nCohen, J. (1978). Partialed products are interactions;\r\npartialed powers are curve components. Psychological\r\nBulletin, 85(4), 858-866.\r\nCohen, J. (1988).Statistical power analysis for the\r\nbehavioral sciences. Hillsdale, NJ: Eribaum.\r\nFalenchuk, O. (2006). A study of unidimensional IRT\r\nmodels for items scored in multiple ordered response\r\ncategories. (Doctoral dissertation Ph.D.), University\r\nof Toronto (Canada), Ann Arbor. Retrieved from\r\nhttp://search.proquest.com/docview/304929555?\r\naccountid=10067 ProQuest Dissertations & Theses A&I\r\ndatabase. (304929555)\r\nFox, J., Nie, Z., Byrnes, J., Culbertson, M., DebRoy, S.,\r\nFriendly, M., . . . Monette, G. (2017). Package ‘sem’.\r\nRetrieved from https://cran.r-\r\nproject.org/web/packages/lavaan/lavaan.pdf\r\nGarson, G. D. (2016). Partial least squares: regression\r\nand structural equation models. Asheboro, NC:\r\nStatistical Publishing Associates.\r\nGoodhue, D. L., Lewis, W., & Thompson, R. (2012). Does\r\npls have advantages for small sample size or non-normal\r\ndata? Mis Quarterly, 36(3), 981-1001.\r\nGordon, M. K. (2016). Achievement Scripts: Media\r\nInfluences on Black Students` Academic Performance,\r\nSelf-Perceptions, and Career Interests. Journal of\r\nBlack Psychology, 42(3), 195-220.\r\nHarwell, M., Stone, C., Hsu, T. & Kirisci, L. (1996).\r\nMonte Carlo studies in item response theory. Applied\r\nPsychological Measurement, 20,101-125.\r\nHair, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt,\r\nM. (2014). A primer on partial least squares structural\r\nequations modeling (PLS-SEM). Thousand Oaks: SAGE\r\nPublications.\r\nHair, J. F., Ringle, C. M., & Sarstedt, M. (2011). PLS-\r\nSEM: Indeed a silver bullet. Journal of Marketing\r\nTheory and Practice, 19(2), 139-152.\r\nHenseler, J., & Chin, W. W. (2010). A comparison of\r\napproaches for the analysis of interaction effects\r\nbetween latent variables using partial least squares\r\npath modeling. Structural Equation Modeling: A\r\nMultidisciplinary Journal, 17(1), 82-109.\r\nHenseler, J., & Fassott, G. (2010). Testing moderating\r\neffects in pls path models: an illustration of\r\navailable procedures. In V. Esposito Vinzi, W. W. Chin,\r\nJ. Henseler, & H. Wang (Eds.), Handbook of partial\r\nleast squares: Concepts, methods and applications. New\r\nYork : Springer .\r\nHu, L. T., & Bentler, P. M. (1999). Cutoff criteria for\r\nfit inde xes in covariance structure analysis:\r\ncriteria versus new alternatives. Structural Equation\r\nModeling: A Multidisciplinary Journal, 6(1), 1-55.\r\nHu, L. T., Bentler, P. M., & Kano, Y. (1992). Can test\r\nstatistics in covariance structure-analysis be trusted.\r\nPsychological Bulletin, 112(2), 351-362.\r\nJöreskog, K. G., & Wallentin, F. Y. (1996). Nonlinear\r\nstructural equation models: The Kenny-Judd model with\r\ninteraction effects. In G. A. Marcoulides & R. E. Schumacker (Eds.), Advanced structural equation modeling\r\n(pp. 57-89). Mahwh, NJ: Lawrence Erlbaum\r\nJöreskog, K. G., Cudeck, R., Du Toit, S. H. C., & Sörbom,\r\nD. (2001). Structural equation modeling, present and\r\nfuture : A festschrift in honor of Karl Jöreskog.\r\nLincolnwood, IL: Scientific Software International.\r\nKaplan, D. (2009). Structural equation modeling :\r\nFoundations and extensions (2nd ed.. ed.). Los Angeles:\r\nLos Angeles : SAGE.\r\nKenny, D. A., & Judd, C. M. (1984). Estimating the\r\nnonlinear and interactive effects of latent-variables.\r\nPsychological Bulletin, 96(1), 201-210.\r\nKlein, A., & Moosbrugger, H. (2000). Maximum likelihood\r\nestimation of latent interaction effects with the LMS\r\nmethod. Psychometrika, 65(4), 457-474.\r\nKraemer, H. C., & Blasey, C. M. (2004). Centring in\r\nregression analyses: a strategy to prevent errors in\r\nstatistical inference. International Journal of Methods\r\nin Psychiatric Research, 13(3), 141-151.\r\nLeite, W. L., & Zuo, Y. Z. (2011). Modeling Latent\r\nInteractions at Level 2 in Multilevel Structural\r\nEquation Models: An Evaluation of Mean-Centered and\r\nResidual-Centered Unconstrained Approaches. Structural\r\nEquation Modeling: A Multidisciplinary Journal, 18(3),\r\n449-464.\r\nLittle, T. D., Bovaird, J. A., & Widaman, K. F. (2006).\r\nOn the merits of orthogonalizing powered and product\r\nterms: Implications for modeling interactions among\r\nlatent variables. Structural Equation Modeling: A\r\nMultidisciplinary Journal, 13(4), 497-519.\r\nLohmöller, J.-B. (1989). Latent variable path modeling\r\nwith partial least squares. Heidelberg, Germany:\r\nPhysica-Verlag.\r\nMarquardt, D. W. (1980). You should standardize the\r\npredictor variables in your regression model. Journal\r\nof the American Statistics Association, 75, 87-91.\r\nMarsh, H. W., Wen, Z. L., & Hau, K. T. (2004). Structural\r\nequation models of latent interactions: Evaluation of\r\nalternative estimation strategies and indicator\r\nconstruction. Psychological Methods, 9(3), 275-300.\r\nMartínez-Ruiz, A., & Aluja-Banet, T. (2013). Two-step PLS\r\npath modeling mode B: Nonlinear and interaction effects\r\nbetween formative constructs. In H. A. W. W. Chin, V.\r\nE. Vinzi, G. Russolillo, & L. Trinchera (Eds.), New\r\nperspectives in partial least squares and related\r\nmethods. Springer: New York.\r\nMonecke, A. (2013). Package ‘semPLS’. Retrieved from\r\nhttps://cran.r-\r\nproject.org/web/packages/semPLS/index.html\r\nMoulder, B. C., & Algina, J. (2002). Comparison of\r\nmethods for estimating and testing latent variable\r\ninteractions. Structural Equation Modeling, 9(1), 1-19.\r\nMueller, R. O. (1996). Basic principles of structural\r\nequation modeling : An introduction to LISREL and EQS.\r\nNew York: Springer.\r\nMuthén, L. K., & Muthén, B. O. (2017). Mplus user`s\r\nguide: Version 8. Los Angeles, CA: Muthen & Muthen.\r\nPark, H. (2000). Comparison of IRT models for ordered\r\npolytomous response data(Order No. 9983591). Available\r\nfrom ProQuest Dissertations & Theses A&I. (304612642).\r\nRetrieved from\r\nhttp://search.proquest.com/docview/304612642?\r\naccountid=10067\r\nQuintana, S. M., & Maxwell, S. E. (1999). Implications of\r\nrecent developments in structural equation modeling for\r\ncounseling psychology. Counseling Psychologist, 27(4),\r\n485-527.\r\nQureshi, I., & Compeau, D. (2009). Assessing between-\r\ngroup differences in information systems research: A\r\ncomparison of covariance- and component-based sem. Mis Quarterly, 33(1), 197-214.\r\nReinartz, W., Haenlein, M., & Henseler, J. (2009). An\r\nempirical comparison of the efficacy of covariance-\r\nbased and variance-based SEM. International Journal of\r\nResearch in Marketing, 26(4), 332-344.\r\nRossee, Y. (2012). lavaan: An R package for structural\r\nequation modeling. Journal of Statistical Software,\r\n48(2), 1-36.\r\nRosseel, Y., Oberski, D., Byrnes, J., Vanbrabant, L.,\r\nSavalei, V., Merkle, E., . . . Barendse, M. (2015).\r\nPackage ‘lavaan’. Retrieved from https://cran.r-\r\nproject.org/web/packages/lavaan/lavaan.pdf\r\nSchumacker, R. E. (2002). Latent variable interaction\r\nmodeling. Structural Equation Modeling, 9(1), 40-54.\r\nSchumacker, R. E. (2016). A beginner`s guide to\r\nstructural equation modeling. New York : Routledge.\r\nSharma, P. N., & Kim, K. H. (2013). A comparison of PLS\r\nand ML bootstrapping techniques in SEM: A Monte Carlo\r\nStudy. In H. Abdi, W. W. Chin, V. Esposito Vinzi, G.\r\nRussolillo, & L. Trinchera (Eds.), New perspectives in\r\npartial least squares and related methods (pp. 201-\r\n208). New York, NY: Springer New York.\r\nWall, M. M., & Amemiya, Y. (2001). Generalized appended\r\nproduct indicator procedure for nonlinear structural\r\nequation analysis. Journal of Educational and\r\nBehavioral Statistics, 26(1), 1-29.\r\nWang, L. J., & Preacher, K. J. (2015). Moderated\r\nmediation analysis using Bayesian methods. Structural\r\nEquation Modeling: A Multidisciplinary Journal, 22(2),\r\n249-263.\r\nWen, Z. L., Marsh, H. W., & Hau, K. T. (2010). Structural\r\nequation models of latent interactions: An appropriate\r\nstandardized solution and its scale-free properties.\r\nStructural Equation Modeling: A Multidisciplinary\r\nJournal, 17(1), 1-22.\r\nYang-Wallentin, F. (2001). Comparisons of the ML and TSLS\r\nestimators for the Kenny-Judd. In D. Sörbom, S. H. C.\r\nDu Toit, & R. Cudeck (Eds.), Structural equation\r\nmodeling present and future : A festschrift in honor of\r\nKarl Jöreskog (pp. 425-442). Lincolnwood, IL Scientific\r\nSoftware International.
描述: 博士
國立政治大學
教育學系
102152502
資料來源: http://thesis.lib.nccu.edu.tw/record/#G1021525021
資料類型: thesis
Appears in Collections:學位論文

Files in This Item:
File SizeFormat
李昭鋆.pdf1.22 MBAdobe PDF2View/Open
Show full item record

Google ScholarTM

Check

Altmetric

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.