A Team of Researchers Led by Prof. Kim Su-young Publishes a Paper in the World-Renowned SSCI Journal Structural Equation
A Team of Researchers Led by Prof. Kim Su-young Publishes a Paper in the World-Renowned SSCI Journal Structural Equation Modeling
Prof. Kim Su-young (psychology, corresponding author) and his teaching student Jeon Min-jeong (lead author) have explained issues regarding estimation accuracy and scaling-related interpretation of the second-order latent growth model through a paper titled Performance of Second-Order Latent Growth Model Under Partial Longitudinal Measurement Invariance: A Comparison of Two Scaling Approaches. These factors had remained unexplained under social science research methodologies due to their complexity. The paper was published in Structural Equation Modeling, an internationally renowned SSCI academic journal in the field of social science research methodologies including psychology, sociology, and education that has been ranked first place on four occasions over the last decade in the fields of Social Sciences and Mathematical Methods as the most frequently cited journal.
A core aspect of social science research is collecting longitudinal data across various points to explain causal relationships that cannot be explained with cross-sectional data and changes in the target variable over time. The latent growth model, frequently adapted for this purpose, is an excellent statistical model using the framework of a structural equation model when analyzing changing patterns over time, which has been widely used in academic fields such as psychology, education, nursing science, and business administration. However, it entails a downside that, in the research of human psychology, it is inevitable for measurement errors to occur and these errors cannot be controlled. On the other hand, the second-order latent growth model allows researchers to control measurement errors that arise during the psychological measurement process of longitudinal data, gaining significant attention in various fields where a psychological construct is applied. Nevertheless, despite these benefits, the scaling of the second-order latent growth model has not yet been fully investigated.
Against this backdrop, the latest research revealed which method, out of the method of using long-applied marker variables or the newly-emerging method of effects coding, leads researchers to a more accurate estimate and connotes higher potential for meaningful interpretations through formula expansion and computer simulation. A team of researchers at Ewha adopted a simulation method that created virtual longitudinal data with programming and conducted model estimations using both the marker variables and effects coding methods before comparing the margins of error. Consequently, it was found that the two methods each showed advantages in their respective areas, but in general, the newer method of effects coding provided more accurate data in a realistic data collection environment and was proven to be a scaling method with a lower probability of errors.
Considering the majority of published dissertations using the second-order longitudinal measurement model are based on marker variables, Prof. Kim and his research team’s paper is the first study that has proven the superiority of the effects coding method. Prof. Kim stated that “Our research outcomes will provide researchers who are planning to analyze longitudinal data by using the second-order longitudinal measurement model and interpret their outcomes with meaningful guidelines and I expect that both domestic and international research methods of longitudinal data will take a leap forward in the future.”