葛虹行为导向AMOS+PLSmodelResearchMethods8——PLS各种分析.ppt

2019/8/7,1,Research Methods in Information Systems,Lecture 8 Path Analysis and Interaction Analysis Using PLS,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,2,Agenda,Introduction to PLS Refining Research Model Interaction Analysis,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,3,Introduction to PLS,Based on Chin,1998, Gefen et al. 2000, and Temme et al. 2006,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,4,Introduction to PLS,Dr. Qing Hu and Dr. Hong Ge,What is PLS? Partial Least Square (PLS) is a statistical method for conducting Structural Equation Modeling (SEM) PLS is a component-based tools emphasize the amount of variance in the data that have be explained by the model (e.g., R2) Thus, parameter estimates in PLS are obtained based on their ability to minimize the residual variances of dependent variables (both latent and observed). This is in contrast with covariance based methods where the objective is to minimize the difference in covariance between the estimated model and the actual data.,2019/8/7,5,Introduction to PLS,Dr. Qing Hu and Dr. Hong Ge,The main benefits of PLS Minimal demands on measurement scales (i.e., measures need not to be interval or ratio) Demand for relatively smaller sample size No requirement for normality in residual distributions. Can be used for theory confirmation, can also be used to suggest where relationships might or might not exist and to suggest propositions for testing later. Avoids two serious problems in covariance-based methods: inadmissible solutions and factor indeterminacy,2019/8/7,6,The history of PLS The PLS approach has its origins back in 1966 when Herman Wold presented two iterative procedures using least squares (LS) estimation for single- and multi- component models and for canonical correlation. The basic PLS design was completed in 1977 by Wold and has subsequently been extended in various ways by Lohmoller (1984, 1989) and Hui (1978, 1982). Wynne Chin of University of Huston is the most significant scholar in IS area who developed PLS-Graph software and promoted its use in IS research.,Introduction to PLS,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,7,Introduction to PLS,LISREL/AMOS vs. PLS,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,8,Introduction to PLS,Different PLS software package,Dr. Qing Hu and Dr. Hong Ge,,2019/8/7,9,SmartPLS Interface,Dr. Qing Hu and Dr. Hong Ge,Introduction to SmartPLS,2019/8/7,10,Introduction to SmartPLS,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,11,,Dr. Qing Hu and Dr. Hong Ge,,2019/8/7,12,Data Analysis Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,,2019/8/7,13,Data Analysis Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,No information of significance,2019/8/7,14,Data Analysis Using SmartPLS,The resampling methods in PLS One of the appealing features of PLS path modeling is that it does not rest on any distributional assumptions Thus, significance levels for the parameter estimates which are based on normal theory are not suitable. Statistics about the variability of the parameter estimates and hence their significance has to be generated by means of resampling procedures.,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,15,The Bootstrapping resampling methods All recent PLS software packages include a bootstrap option. The bootstrap procedure approximates the sampling distribution of an estimator by resampling with replacement from the original sample. In order to derive valid standard errors or t-values, applying bootstrapping is superior to the other resampling methods. Bootstrapping is considered superior over the other two resampling techniques: Blindfolding Jackknifing,Data Analysis Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,,2019/8/7,16,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,Setting Bootstrapping parameters in SmartPLS,2019/8/7,17,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,18,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,19,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,20,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,21,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,,,,2019/8/7,22,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,23,Dr. Qing Hu and Dr. Hong Ge,Bootstrapping Using SmartPLS,2019/8/7,24,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,25,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,Cases 200, Sample 100,2019/8/7,26,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,Cases 200, Sample 200,2019/8/7,27,Bootstrapping Using SmartPLS,Dr. Qing Hu and Dr. Hong Ge,Cases 73, Sample 200,2019/8/7,28,Refining Research Model,Dr. Qing Hu and Dr. Hong Ge,2019/8/7,29,Refining Research Model,Combining AMOS with PLS AMOS provide advanced statistics such as Modification Indices that allow researcher to test alternative measurement models and path models. PLS provide extensive statistics and convenience for reporting model fit. By combining these two techniques, we can take the advantage of both packages Use AMOS to discover the mode。