Dear Dinesh Kumar

Before answering your questions on what are the best practices for assessing model fit in SmartPLS, you need to know what approach you are using for your research, it is CB-SEM or VB-SEM.

If your answer is VB-SEM, then, goodness of fit is irrelevant in this context.

You can also attempt by using PLSc which mimics CB-SEM.

Most importantly, you need to know which approach and then the requirement of that particular approach whether data normality is a must or not. Then only deal with the issue one by one.

Hope it helps.

Best regards

Low

Dear Dinesh Kumar

Assessing the adequacy of a model's fit in SmartPLS can present challenges when the data may violate the normality assumption, for instance, due to outliers, skewness, or non-normal distributions. Nonetheless, a number of effective approaches can aid in evaluating the model fit in such scenarios:

• Utilize robust estimation techniques: SmartPLS offers a range of robust estimation methods that are less sensitive to deviations from normality assumptions, such as the Partial Least Squares Bootstrap (PLS-Bootstrap) and Weighted Least Squares (WLS) methods, which can be more appropriate when non-normal data or outliers are present.
• Examine the data distribution: Before evaluating model fit, examining the data distribution is crucial. Descriptive statistics such as skewness and kurtosis can assist in evaluating normality assumptions. In case the data is skewed or non-normal, non-parametric techniques or data transformations may be necessary prior to conducting the analysis.
• Utilize multiple fit indices: It is important to use multiple fit indices to evaluate model fit. SmartPLS offers various fit indices, such as R-squared, Q2, Goodness of Fit Index (GoF), and standardized root mean square residual (SRMR). This strategy can provide a comprehensive evaluation of model fit and aid in detecting potential issues with the model.
• Assess coefficient significance: Examining coefficient significance is also important when evaluating model fit. Confidence intervals can be obtained using the bootstrapping method and statistical significance of the coefficients can be tested. If the coefficients are not statistically significant, this could indicate a poor fit between the model and the data.
• Check for outliers: Outliers can significantly affect model fit. It is important to examine the data for outliers and consider removing them from the analysis if they are influential. SmartPLS provides several diagnostic tools such as leverage plots and Cook's distance, that can help identify potential outliers.

Overall, when handling non-normal data, outliers, or skewness in SmartPLS, it is important to use robust estimation methods, examine data distribution, utilize multiple fit indices, assess coefficient significance, and check for outliers. Employing these best practices can lead to a more accurate evaluation of model fit and detection of potential model issues.

Hope it helps.

Best regards

Kadhim