Empirical Study About Index Futures Impact To Stock Market Based On Variables Selection

In this paper, the variable selection method is applied in the study of select a part of the stocks from CS7I00 stocks, to establish a stock portfolio which operated simultaneously with GSI300 stock index futures, in order to provide a new option for investors in stock portfolio construction methods. In this paper, the method to construct stock portfolio is substantially different with the previous method based on experience or subjective judgement. As used herein, the method is based on the characteristic of the CSI300 stock index futures and CSI300 stock data, using the mathematical statistics variables selection models, thus analysis the problem more objectively, to select the more suitable stock portfolio to hedging or arbitrage using CSZ300 stock index futures. In this paper, the minutes highly frequency data of GSI300 stock index futures and CSI300 stock on September 25,2014 is selected as object in empirical study, ten variable selection methods are applied in the study of selecting portion of stocks from the CSI300 stocks, then in accordance to the R value obtained from the final regression model established based on the selected stocks from CSI300 index futures, the model selection effects were arrayed, It is found that the number and precision are different according to the selected ways, meanwhile the regression determination coefficient R values from stepwise regression and the combination both are almost 0.64. It is not only shown that the methods have positive effect on the operation of the futures market and obtaining benefit by stocks investment, but also indicated that the method is new breakthroughs to the relevance previous research emphasis on subjectivity and empirical.According to the research process and findings, the following conclusions are concluded in this paper:1. In the study of specific examples in this paper, the effects are compared as follows: Stepwise Regression+Lasso> Lasso> weighted Lasso> weighted Group-Lasso> SIS-Lasso= SIS-weighted-Lasso, to note that there are two ways due to the limitations of the method, failed to make the variable selection results.2. In the high-dimensional variable selection, the traditional AIC, BIC criterion has limitations. In the empirical study, through the analysis in running program, it is found that the reason why Lasso, Group-Lasso cannot directly choose a good stock portfolio, is that the AIC, BIC and other criteria in model cannot achieve optimal value at the optimum place, therefore, how to improve or set high dimensional variable selection model selection criterion is an issue to be further studied.3. From the perspective of variable selection, stock selection is feasible by statistical methods. In previous studies, the method to select stock portfolio is more subjective, not based on the perspective of data features analysis to choose the stock to construct the stock portfolio, in this research, the variable selection method is applied to construct stock portfolio successfully.