How to make a quantitative social science research
Author:Li Yunzhen Date:Jan 4, 2017 Clicks:

In mid-November, Associate Professor Sun Yongqiang gave a lecture on “structural equation modeling—quantitative analysis in social science research” in WHU School of Information Management (SIM).

Professor Sun first introduced the appearance and development of Structural Equation Modeling (SEM). SEM is now a widely applied strategy in social science researches. It was well established in the 1980s but rarely known and applied in China. “In the realm of social science, which includes economy, marketing and management, multiple causes and effects are prevalent in phenomena, and usually there would be variables that cannot be observed directly.  This complex situation calls for something beyond the traditional statistics method. In 1980s, the introduction of SEM solved this problem.”

According to Professor Sun, SEM has five main advantages compared with traditional methods:

First and foremost, SEM could deal with multivariate data. In traditional regression analysis and path analysis, variables are calculated separately. Existence and interaction of other variables are not taken into consideration. Secondly, the SEM allows for acceptable measurement errors. Variables such as attitude and behavior usually cause errors in the measurement. Such errors are simplified or neglected in traditional methods as those mathematical methods prefer a perfect condition. Next, SEM allows researchers to simultaneously evaluate the structure of determiners and the relationships between them. Calculating the relation between subjects and determiners and evaluating the interdependency among determiners used to be two separated steps; while with SEM, these two steps are considered concurrently. Fourthly, SEM deals with an observational model with higher flexibility. Traditional factor analysis can hardly deal with an index affected by multiple determiners. For example, if a student is asked to do a math test in English, his score will not only be affected by his or her mathematical skills, but also the English proficiency of this individual. Finally, the model’s overall goodness of fit of same example data can be estimated in SEM. Under the SEM approach, not only can the relationship among variables be estimated, the overall goodness of fit of same sample data with different models can also be determined. Therefore, the researchers can find the most appropriate model for their sample data.

Furthermore, Professor Sun also emphasized the importance of methodology and the philosophy of science. He explained the difference between natural science and social science research: In social science, the society we study is a much more complicated system compared with the perfect, idealized hypothesis of natural science. The results of a case are not definitely caused by the observable factors. Hence, it only explains a possibility rather than a certainty.

Organized by the postgraduate student union of SIM, this lecture was a part of the 11th Science and Technology Festival of Wuhan University. During this festival, students from different colleges of WHU would organize various activities, such as quizzes, interdisciplinary seminars or lectures and other related recreational and sports activities. Usually held in November, this festival has become an important event for students, especially for postgraduates and doctors, to have a break from their hard study.

Edited by Tang Yedan, Wu Siying & Li Yunzhen

In mid-November, Associate Professor Sun Yongqiang gave a lecture on “structural equation modeling—quantitative analysis in social science research” in WHU School of Information Management (SIM).

Professor Sun first introduced the appearance and development of Structural Equation Modeling (SEM). SEM is now a widely applied strategy in social science researches. It was well established in the 1980s but rarely known and applied in China. “In the realm of social science, which includes economy, marketing and management, multiple causes and effects are prevalent in phenomena, and usually there would be variables that cannot be observed directly.  This complex situation calls for something beyond the traditional statistics method. In 1980s, the introduction of SEM solved this problem.”

According to Professor Sun, SEM has five main advantages compared with traditional methods:

First and foremost, SEM could deal with multivariate data. In traditional regression analysis and path analysis, variables are calculated separately. Existence and interaction of other variables are not taken into consideration. Secondly, the SEM allows for acceptable measurement errors. Variables such as attitude and behavior usually cause errors in the measurement. Such errors are simplified or neglected in traditional methods as those mathematical methods prefer a perfect condition. Next, SEM allows researchers to simultaneously evaluate the structure of determiners and the relationships between them. Calculating the relation between subjects and determiners and evaluating the interdependency among determiners used to be two separated steps; while with SEM, these two steps are considered concurrently. Fourthly, SEM deals with an observational model with higher flexibility. Traditional factor analysis can hardly deal with an index affected by multiple determiners. For example, if a student is asked to do a math test in English, his score will not only be affected by his or her mathematical skills, but also the English proficiency of this individual. Finally, the model’s overall goodness of fit of same example data can be estimated in SEM. Under the SEM approach, not only can the relationship among variables be estimated, the overall goodness of fit of same sample data with different models can also be determined. Therefore, the researchers can find the most appropriate model for their sample data.

Furthermore, Professor Sun also emphasized the importance of methodology and the philosophy of science. He explained the difference between natural science and social science research: In social science, the society we study is a much more complicated system compared with the perfect, idealized hypothesis of natural science. The results of a case are not definitely caused by the observable factors. Hence, it only explains a possibility rather than a certainty.

Organized by the postgraduate student union of SIM, this lecture was a part of the 11th Science and Technology Festival of Wuhan University. During this festival, students from different colleges of WHU would organize various activities, such as quizzes, interdisciplinary seminars or lectures and other related recreational and sports activities. Usually held in November, this festival has become an important event for students, especially for postgraduates and doctors, to have a break from their hard study.

Edited by Tang Yedan, Wu Siying & Li Yunzhen

关闭

Copyright @ 2014 Wuhan University