The Merging Method Of Multilevel Data And It’s Application

As the situation that the dependent variable data of layer structure having a small sample of data is not suit for Hierarchical linear modeling to analyze,this paper innovates a new way by merging the raw data to expand sample size,focusing on establishing cross sectional data Hierarchical Liner modeling by merging data based on same organization, same size,same area;establishing continuous dependent variables panel data Hierarchical Liner modeling by merging data based on the numerical size,the correlation between independent variable and dependent variable,the principal component analysis respectively.Establishing binary discrete dependent variable data and multinomial discrete dependent variable data Hierarchical Liner modeling by merging data based on same organization,same size,same area.and then giving the corresponding relationship between fitted values and raw data.Applying the Hierarchical Liner modeling based on merging method to the actual data and simulation data in Hainan Province,Fujian Province,Guangdong Province,Guangxi Province,Yunnan Province,and the data is collected by Hierarchical Liner modeling research group of Hainan university’s school of economics and management.This paper establishes merge-cross sectional data Hierarchical linear modeling based on same district,in which dependent variable is output per acre,the independent variables are acreage,chemical fertilizer,human capital in level one,and urbanization rate,GDP per head,professional management staff number in level two;merge-panel simulations Hierarchical linear modeling in which dependent variable is output per acre,independent variables are acreage,human capital,chemical fertilizer in level one,and professional management staff number in level two,The loan share of income in level two;merge-panel actual data Hierarchical linear modeling in which dependent variable is output per acre,independent variables are acreage,human capital in level one,and agricultural investment proportion, GDP per head in level two;merge-logistic Hierarchical linear modeling in which dependent variable is incidence ratio of professional association,independent variables is packaging technology in level one,urbanization rate in level two.merge-logit Hierarchical linear modeling in which dependent variable is packaging technology,independent variables is acreage in level one.and GDP per head in level two.Through the practical application of the banana industry,the principal conclusions of this analysis are as follows:The banana-growing region in five southern provinces can raise output per acre by expanding the planting scale,raising the number of regional professional managers and the agricultural investment proportion will promote this relationship.In the situation that whether or not join the professional association,it comes out that the the growers in five southern provinces will prefer to joining the association when they choice biological pesticide,and the increasing urbanization rate will promote this relationship.The conclusion of selecting the fresh-keeping packaging technology in five southern provinces is the growers will prefer to the cardboard boxes packaging technology when they expanding the planting scale,and the increasing local GDP per head will promote this relationship.The MAPE in three aspects applications of the merge-banana continuous independent variable Multilevel model are0.9904%,3.7708%,6.2219%.It means the precision of the model is very high,and the results of the empirical analysis is very much in line with reality,all of this fully illustrated that the merge-Hierarchical linear modeling has high practical value.In view of the above conclusions,in order to achieve the goal of the banana industry scientific and efficient development,the government should focus on improving local credit,increasing government input and further advancing the urbanization.