*Model 5: Evaluation Model*

The data collection phase involves more than two periods of time with the same questionnaire. For the measurement model, the constructs will be evaluated according to the time. While in the structural model, the analysis starts by developing one PLS model for each period of time. In this stage, the direct effect and carry-over effect will be examined based on the path analysis. Then, to test the changes in the path coefficient over time, the bias-corrected confidence interval is computed. Next, paired *t*-test of the changes in the level of the construct over time is computed.

#### **5. Discussion**

Structural Equation Modeling (SEM) is one of the flexible methods for analyzing survey data. This method is used as a statistical tool for evaluating the relation between latent and observed variables [73]. SEM can be defined as a combination of several multivariate analysis techniques [74], such as path analysis [75] and the common factor or latent variable model [76]. Thus, this study reviewed the methodology that used SEM as a base framework for analyzing panel survey data. There are two methods that have been discovered in this study which are LGCM and several approaches in the PLS-SEM. The trend of publications related to the panel survey data is increasing over the year. The findings show that the application of LGCM is preferable compared to the PLS-SEM in analyzing the panel survey data. We can see the pattern in the bibliometric analysis where the findings are dominated by the LGCM. This is because the ability and flexibility of LGCM in handling panel survey data are better than the PLS-SEM. Among the ability of LGCM, it can describe the developmental trajectory of a single person and capture individual variations over time. In other words, this method can assess the changes in intra-individual (within the individual) as well as inter-individual (between individuals) variation. LGCM can also identify the important predictor variables that contribute to the individual's growth change over time. [77] described the several advantages of LGCM which permits the investigation of inter-individual differences in change over time and allows the researcher to investigate the antecedents and consequences of change. LGCM also provides group-level statistics such as mean growth rate and mean intercept, can test hypotheses about specific trajectories, and allows the incorporation of both time-varying and time-invariant covariates. This could be the main reason why LGCM is preferable compared to the PLS-SEM, even though it has several limitations. According to [22], the existing approaches of PLS-SEM in panel survey data still have limitations. The approaches also show a lack of flexibility in analyzing the panel survey data in one structural model. Thus, this study employed content analysis to identify the existing approaches of PLS-SEM in analyzing panel survey data and its limitations.

The findings show that there are five existing approaches of PLS-SEM that have been used in analyzing panel survey data. Among the existing approaches of PLS-SEM, the Evaluation approach is the most flexible approach in analyzing panel survey data. Thus, this study discussed this approach more than the other four approaches. This approach consists of three stages in analyzing the panel survey data. The analysis measured the direct effect and the special effect which is the carry-over-effect. Carry-over-effects are effects from one construct at one point in time to the same construct at a subsequent point in time [78]. In stage one, the direct effect and carry-over-effect are assessed by estimating the single PLS model separately across the time. With this, the separate direct effect between the endogenous and exogenous predictors across time can be assessed. Hence, one structural model to access the whole changes (trajectories) and the factors that influence those changes simultaneously cannot be established. In the second stage, the multi-group analysis is employed to assess the strength of direct effects and the carry-over-effects over time. This strength is measured by the changes in the size of the path coefficient and bias-corrected confidence interval. The limitation at this stage is that the factors that influence those changes in the carry-over-effect simultaneously in one structure modal cannot be measured. In the last stage, paired t-test is employed to assess the mean difference between the constructs. The limitation in this stage is that only the mean difference for two points at a time for each construct can be assessed. In addition, the paired t-test requires a few assumptions and the most concern for the researcher is the distributional assumption. Hence, all these stages in the Model 5 approach do not have one structural model to represent the whole changes in the repeated measure. In addition, current approaches cannot capture the individual trajectory, mean of the trajectory of the sample or entire group, the evaluation of individual differences in trajectories, and assess the potential incorporation of predictors of individual differences in trajectories. Consequently, with all these limitations, PLS-SEM is less frequently used for analyzing panel survey data.

#### **6. Conclusions**

In conclusion, this study explored the distributions and trends of publications related to the panel survey data. This study also explored the trends of publications according to the Latent Growth Curve Model (LGGM) and PLS-SEM in analyzing the panel survey data. The records were retrieved from the bibliographic databases of Scopus and Web of Science Core Collection (WoSCC). The trends of publications related to the panel survey data showed an increasing trend. However, in the context of the statistical method, the LGCM is preferable compared to the PLS-SEM in analyzing the panel survey data, even though the LGCM has several limitations as highlighted in previous studies. This is because the PLS-SEM shows a lack of capability in handling panel survey data, even though it has five different approaches in analyzing them. The most flexible approach of the PLS-SEM in handling panel survey data is model 5 since it can measure the direct effect, carry-over effect, and the changes of path coefficients over time. However, based on the review, this approach still has some space for improvement. This method cannot capture an individual trajectory, the mean of the trajectory of the sample or entire group, the evaluation of individual differences in trajectories, and assess the potential incorporation of predictors of individual differences in trajectories. Besides, these current approaches are not as flexible as LGCM since it has the ability to use variables simultaneously as independent and dependent in the same model. Therefore, this systematic review could help researchers choose a more suitable method to analyze panel survey data.

**Author Contributions:** Z.M.G.: Conceptualization, Data Collection, Methodology, Formal Analysis, Visualization, Preparation of original draft. W.F.W.Y.: Conceptualization, Methodology, Writing, Review & Editing. W.M.W.O.: Methodology, Review & Editing. All authors have read and agreed to the published version of the manuscript.

**Funding:** The authors would like to thank Universiti Teknologi MARA (UiTM) for funding this research under PYPA.

**Acknowledgments:** The authors would like to thank the reviewers for their helpful and constructive comments and suggestions that greatly contributed to the improvements of the final version of this paper.

**Conflicts of Interest:** The authors declare no conflict of interest.
