More info
Taking into account the structural equation modeling with partial least squares (PLS-SEM) first by birth (Wold) in 1974 expressed, then Lmvlr (Lohmoller) and in 1989 would deal with a more developed approach it presented. Taking into account the structural equation modeling with partial least squares (PLS-SEM) is the second generation of structural equation modeling and compared to the last generation that covariance-based methods (Covariance Based) were great advantages are as follows. Be statistically sound test with small sample sizes, or with non-normal data, performance measurement models of the manufacturer, power and predictive power superior to the previous generation, the ability to deploy ever more sophisticated models, research, exploration, development theory and theory, theory and hypothesis testing, testing assumptions include mediating variables, use variables classified, check convergence.
The first generation of structural equation modeling (Covariance-based SEM techniques)
These methods - methods of covariance - by Jvrsgvk (Joreskog) are introduced in 1969. Its main purpose was to validate models, and high-volume sample needed for this work. To estimate the path and again using to minimize the difference between the matrix of variance - covariance observed and projected payments. The most widely used approach to calculating the coefficients in the first generation, the maximum likelihood estimation approach requires data on the observed variables or questionnaire that these variables must be sure that they follow a normal distribution. And the most famous Nrafzarhay this group (Amos, EQS, Lisrel) are.
Second-generation methods of structural equation modeling (Component based SEM techniques)
Component-based approaches (Component based), which later Partial Least Square method (Partial Least Squares) changed its name, was coined in 1974 by the son. PLS approach (PLS) is composed of two main phases: a review of the measurement model, structural model and the overall model b: Examining the relationships between structures. Review articles and studies published in the last decade, the growing extent of use of the PLS (PLS) for statistical analysis and thesis papers and so on. That the above mentioned reasons, the main reason for the superiority of this method for small samples listed next because of problems related to non-normality of the data that researchers are faced in some studies. Finally, the last reason to use this method of dealing with measurement models of the manufacturer (Formative) is. In the first generation structural equation modeling with software including (Amos, EQS, Lisrel) was implemented, requiring large sample size. For example Bvmsma and Hooglund (Boomsma & hoogland (2001)) 200 and above to run models, structural equation by the appropriate software know as PLS (PLS) can run the model with the number of samples is much less support.
Model evaluation method
The first factor in the assessment model, should be considered, one-dimensionality of the model parameters. This means that each index in the total index, with a load factor, only a latent variable, loaded. For this purpose, the load factor is larger than 60/0. It should be noted that the load factor is smaller than the small 40/0 and should be deleted from the set of indicators. This has to be removed manually with indicators that have less than 40/0 loadings are done (Giffen and Strub, 2005, Grbyk and Anderson, 1988).
Cronbach's alpha coefficient (Cronbach alpha (CA))
Another factor in the assessment of internal consistency reliability (Internal consistency reliability) model, the value of this parameter, variable from 0 to 10, with higher values of 70/0 is accepted and less than 60/0 unfavorably (Cronbach, 1951).
Structural reliability coefficient (Composit Reliability (CR))
Another factor that internal consistency reliability evaluation model. The values of these coefficients, from 0 to 1, with higher values of less than 60/0 70/0 accepted and will be unfavorably (wörthsee et al., 1974) Moreover, to ensure the reliability index (Indicator Reliability), all loads must be greater than the index 70/0 and 05/0, at least at the level of meaningful (China, 1988). For this purpose, the output Bootstrapping analysis and statistic T used in the report output.
Credit convergence
Highly correlated indicators of a correlation between structural structures compared to other states that the model should be evaluated. To evaluate the validity Smart pls convergence in the application of the AVE (Average variance extracted) is used. The values of these coefficients, from 0 to 1, with higher values of 50/0 is accepted (Fronleichnam and Larkr, 1981).
Diagnostic validity
In order to evaluate this type of credit, the output of the Cross loading the software and the criteria Fornell-Larcker (Fronleichnam and Larkr, 1981) can be used. For example, as shown in the figure below, the values of variables .... and ..... and ..... factor loadings respectively 78/0, 67/0 and 63/0 are loads than any other factor in the row for each of these variables are larger. To increase or decrease the accuracy of the numbers presented in the report can also be of potential use Decrease Decimal places and Increase Decimal places. Fornell-Larcker criteria also refer to the fact that the square root of the variance values described (AVE) any structure, larger than the values of its correlation with other structures Bashd.bdyn structural purpose, we use the Latent variable correlations. In this table, the values of the correlation between latent variables or dimensions are shown in the model. It is known that the diagonal of the table, one. Fornell-Larcker to evaluate the criteria necessary to copy and paste the table to the environment, such as Excel transfer. Then, the values on the main diagonal matrix with the square root of the variance values described above AVE that replaced, we finally reach the table below. As mentioned, the values in the diagonal of the matrix, must be greater than all the values in the corresponding column. For example, in the column ....., the value in the diagonal of the matrix, which is larger than the values 72/0, 54 / 0,66 / 0 and 53/0 in the same column. It should be noted that according to available data in columns corresponding to variables .... and ..... Table, Fornell-Larcker criteria have been met.
Structural model evaluation method
The first key criteria used for this purpose in the application SmartPLS, the coefficient of determination is R2. An R2 correlation between the amount of variance explained by a latent variable is measured by the total amount of the variance. The value of this coefficient varies from 0 to 1 that larger quantities, is more favorable. China (1988), values close to optimum 67/0, nearly 190/0 to 33/0 casually and near poor. The figure, from the Report Application for project TEST visible.
Evaluation route between latent variables in the model
At this stage, the researcher must factor algebraic symbols, size and its significance level itself. Routes coefficients signs in the opposite direction of the expected hypothesis, the hypothesis was not confirmed. The size of the coefficient indicates the strength of relationship between two variables lies. Some researchers believe that the path coefficient greater than 100/0 shows a certain amount of influence in the model (Huber et al., 2007). Path, from the Path Coefficient reporting software are visible. In addition, the path must be at least 05/0 meaningful through bootstrapping capabilities on the option (path coefficients (mean, stdev, t values
Effect size
Another factor in assessing the validity of the model is the size of the effect (Effect size) or F2 is Cohen (Cohen, 1988). The size of the effect, it pays to express whether an independent latent variable, a significant impact on the dependent variable or not. The value of the coefficient of determination R2 is calculated. F2 values between 02/0 to 15/0 have little impact indicator, between 15/0 to 35/0 from 35/0 indicative of a larger marker moderate impact and high impact on the dependent variable is the independent variable (Cohen, 1988; Giffen et al. 2000). It should be noted that due to limitations in computing software F2, this parameter, the researcher must be manually calculated for individual communication paths between the latent variables.
The model predictions using acetone Napramtry glycerate
Finally, the prediction model using acetone Napramtry glycerate (glycerate, 1975) measured, the test can be done using the blindfolding (Tnhavs et al., 2005). Q2 evaluates the success of this prediction. Q2 and large positive, demonstrate the ability to predict top model.
References
Hanafizadeh, message and Ravasan Zare, A. (1391), analysis of multilevel structures using software smart pls, cashmere Publications, Tehran.
December, AS. (1381). Causal path analysis and management science. Journal of Higher Education Complex Qom, 4 (15), 59-96.
December, A., et al (1391). Modeling route - Structural management: software application smart pls, Tehran, look at knowledge.
Judge, Ali-Zadeh, A. (1393), structural equation modeling software with PLS, publications (SID), Tehran
Home Site Software Smart pls (http://www.smartpls.de/)