Study On The Modeling Methods Of The In-output Statistical Models In Agriculture

The article established the in-output models of Chinese agriculture by introducing the multivariate methods as follow: Canonical Analysis, Stepwise Regression Analysis (SRA), Factor Regression Analysis (FRA), Principle Component Regression Analysis (PCRA) and so on. It makes comparisons and appraisements about the models at the same time, so as to get a good view of the input and output elements in Chinese agriculture by the aid of quantitative analysis. At the beginning the article selects 3 couples of Canonical Correlation Variables by the multivariable statistical method of Canonical Analysis between two groups of variables, and choose out the most important variables between the two groups—the input group and the output group. Based on the Canonical Analysis the article established the Stepwise Regression Models, regarding those important variables in the input group as dependent variables and those ones in the output group as independent variables. But the results are not ideal. There exists very serous multicollinearity among the dependent variables. We also can see that from the fact that some certain correlation coefficients are negative, which is abnormal phenomenon. To eliminate this kind of phenomenon, the article adopts the FRA and PCRA. This time the results are wonderful and it gets the most appropriate regression models moreover the multicollinearity does not exist any more. The whole analysis presents the real conditions of agriculture in China impersonally and reasonably. At the end the article reaches the conclusion that the Chinese agriculture is still of high degree of consumption, low level of efficiency, and that the irrigating water can not be absorbed efficiently, the funds from the government did not show conspicuous effect, which indicates that the utilizing rate of the funds is very low.