Functional Canonical Correlation Analysis and Its Application in Economic Data Analysis

摘要

Classical canonical correlation analysis can study the correlation relationship between two sets of traditional variables using cross-section data, but it cannot analyze the correlation relationship between two functional variables using panal data. Functional canonical correlation analysis (FCCA) solves the problem. FCCA can find correlative forms and corresponding degree. In the paper the basic idea and method of FCCA is demonstrated, and then it is applied into analyzing covarying modes between annual income per capita and living expenditure of urban households. We find that it has good properties and significance in analying economic data and can mine economic laws which cannot be easily found with traditional methods.