SEM Terms Introduction of ADANCO & SmartPLS.pptx

SEM Terms

1. Structural Equation Modeling (SEM): By ADANCO & SmartPLS Organised By: Prof. Seetharaman, Dean (Academics) and Dr. K. Maddulety, Deputy Director S P Jain School of Global Management Dubai • Mumbai • Singapore • Sydney   www.spjain.org 1

2. Presentation Outline Objectives of the Workshop About SEM and SEM & Other MV Techniques SEM: Nomenclature and Representation. Intro to Confirmatory Factor Analysis (CFA) ADANCO Graphics : Hands On SmartPLS Graphics : Hands On CFA or MM : Reliability and Validity Analysis SEM Result Analysis : ADANCO/SPLS Text Output 2

3. Objectives of the Workshop The workshop primarily has twofold objectives: To introduce the basic concepts related to SEM, and To illustrate the applications of SEM using the ADANCO/ SmartPLS software. 3

4. ABOUT SEM SEM is a multivariate technique that seek to estimate a series of inter-related dependence relationships simultaneously. SEM Vs Other MV Techniques Confirmatory Analysis rather than Exploratory Analysis Estimate Measurement Error Variances . Includes both Unobserved (Latent) and Observed Variables Estimate Complex Relationship (series of dependent –Independent relationship) 4

5. Sridhar Vaithianathan, IBS Hyderabad 5

6. SEM: nomenclature and Representation Latent Variables Vs Observed Variables Exogenous Latent Variables Vs Endogenous Latent Variables Reflective Constructs Vs Formative Constructs (Page 62) Recursive Model Vs Non-recursive Model Measurement Model and Structural Model General Purpose of statistical modeling To test whether fit exist between the hypothesized model and the sample data. (i.e.) Does the Model fits the data??? Data = Model + Residual 6

7. Why covariance analysis ? To eliminate some systematic error outside the control of the researcher that can bias the results. To account for differences in the responses due to unique characteristics of the respondents. E.g.: In testing advertising, effects may differ depending upon the time of the day, or the composition of the audience and their reactions. Moreover the personal differences such as attitudes or opinions, may affect the responses, but the analysis does not include them as treatment factors unlike experimental research. 7

8. 8 Where, PU – Perceived Usefulness PEoU - Perceived Ease of Use

9. REFLECTIVE CONSTRUCTS VS FORMATIVE CONSTRUCTS 9

10. REFLECTIVE CONSTRUCTS Reflective Constructs are based on the assumptions that (1) latent constructs cause the measured variables, (2) the measurement error results in an ability to fully explain these measures.. ( Diamantapolous , 1999; 2009) 10

11. FORMATIVE CONSTRUCTS 11 Formative Constructs are based on the assumptions that (1) the measured variables cause the construct, and (2) the error in measurement is an inability of to fully explain the construct. The construct is not latent in this case. ( Diamantapolous , 1999; 2009)

12. SEM : Measurement Model and Structural Model A conventional model in SEM Terminology consists of two models The Measurement Model and The Structural Model The Measurement model represents how measured variables come together to represent constructs. The Structural model shows how constructs are associated with each other. 12

13. 13 Φ 21 (Phi) λ 11 λ 21 λ 31 λ 12 λ 22 λ 32 ξ 1 - Xi (KSI or KZI) ξ 2 δ 11 δ 21 δ 31 δ 12 δ 22 δ 32 δ 11 δ 21

14. 14 γ 11 - gamma ε 32 δ 11 δ 21 δ 31 ε 12 ε 22 λ 11 λ 21 λ 31 λ 12 λ 22 λ 32 ξ 1 Xi (KSI or KZI ) η 1- Eta ζ zeta

15. Intro to Confirmatory Factor Analysis (CFA) CFA is a way of testing how well measured variables represent a particular construct. Purpose of CFA Confirms a hypothesized factor structure. Used as a validity procedure in measurement research. 15

16. EFA Vs CFA EFA Explores data for patterns. EFA reduces a large number of variables into a smaller and more manageable number of factors. Factors are derived from statistical results, not from theory and hence they can only be named only after performing factor analysis. Whereas in CFA the one needs to specify clearly the factor structure before the results could be computed. CFA tells us how well the specification of the factor structure matches the reality. CFA provides a confirmatory test of our measurement theory. 16

17. 17 E E E E V1 V2 V3 V4 Factor 1 Factor 2 Exploratory Factor Analysis Confirmatory Factor Analysis E E E E V1 V2 V3 V4 Factor 1 Factor 2

18. ADANCO Graphics Icons , Terminology & Model identification 18

19. ADANCO HANDS ON…. Working with ADANCO Graphics LAB Session – Sample Data Bring it on… 19

20. SmartPLS Graphics Icons , Terminology & Model identification 20

21. SmartPLS HANDS ON…. Working with ADANCO Graphics LAB Session – Sample Data Bring it on… 21

22. CFA RESULTs ANALYSIS ADANCO Output Model Variables and Parameter Summary Parameter Estimates Model Fit Summary Model Misspecification 22

23. CFA HANDS ON…. (Continuation..) Reliability and Validity Analysis 23 Content/Face Validity Reliability Convergent Validity Discriminant Validity Nomological Validity ( Examines whether the correlation between the constructs in the measurement theory makes sense.)

24. Reliability: Cronbach’s Alpha Cronbach’ s Alpha is the most prominent reliability coefficient. Dijkstra- Henseler's rho (ρ ) Jöreskog's rho (ρ ) It measures the reliability of a set of indicators. Value ranges between 0 to 1 (if all indicators have positive correlation). Greater than 0.7 acceptable (0.6 accepted for survey research) 24

25. Composite reliability Measures the overall reliability of a set of items loaded on a latent construct. Value ranges between 0 and 1. Value greater than 0.7 reflects good reliability. Between 0.6 – 0.7 is also acceptable if other indicators of the construct’s validity are good (Hair et al. 2006). Note: Higher values of Cronbach’s alpha and composite reliability indicates that the internal consistency exists. 25

26. Indicator reliability Denotes the proportion of indicator variance that is explained by the respective latent variable. It is similar to the idea of communality in EFA. Value ranges between 0 – 1. If indicator and latent variable are standardized, the indicator reliability equals the squared indicator loading. 26

27. Convergent validity The items that are the indicators of a construct should converge or share a high proportion of variance in common, known as convergent validity . Value ranges between 0 – 1. Factor loadings should be greater than equal to 0.5, AVE should be higher than 0.5 ( Fornell & Larcker , 1981). 27

28. Average variance extracted AVE is comparable to the proportion of explained variance in factor analysis. AVE should be calculated for all the latent constructs in the CFA model. Value ranges between 0 – 1. 28

29. Discriminant validity Discriminant validity is the extent to which a construct is truly distinct from other constructs. Fornell & Larcker (1981) Criterion Idea: A latent variable should explain better the variance of its own indicators than the variance of other latent variables. Realization: The AVE of a latent variable should be higher than the squared inter-construct correlations. Cross Loadings The loading of an indicator on its assigned latent variable should be higher than its loadings on all other latent variables. 29

30. structural model testing 30

31. Thank you One and all 31