Applicantions Of Transfer Functon Models In Stock Market Analysis

With the rapid development and improvement of Chinese stock market, there are moreand more researches in stock market by using time series analysis. Transfer function modelshave been widely applied to the fields of economy、industrial production and engineering, butis rare in the field of stock.In the paper, transfer function models are used to analyze Shanghai composite index, wecan get the relationship between price and volume by calculating their CCF in bull、bear andbalance market, and making comparison with HIS、Dow Jones Industrial Average andSinopea. The conclusions are: the relationships have some same points and difference for thesame index under different situations and for different indexes under the same situation.Some are that price and volume are synchronous or affect each other, others are that pricecauses to volume or volume causes to price. Establish transfer function models for outputseries volume, with input series price. The advantage of models is that the transfer functionsection reflects the relationship between price and volume, and noise section reflects the ownrelationship of output series. If time series are highly interrelated with input time series andthe total number of lag phases are more, establishing transfer function models areadvantageous thanARIMAmodel. Otherwise, ARIMAmodel is advantageous.In the empirical research of multivariate transfer function models, we suggest removinginput series which don’t meet the condition by using the stepwise selection and testing theCCF between output series and input series, and then establishing models. When input seriesare highly interrelated, suggest make a new input series with weight. The models set up bythe method is advantageous than the models with highly interrelated input series.In this paper, three new models are proposed for the appearance of intervention events.The transfer function models with intervention input series are applied in empirical analysis,proved that when input series is impacted by intervention, the new models are advantageousthan traditional transfer function models.