The Study Of Medical Time Series/Cross Section Based On Varying-Parameters Models

Objective:The large span time-series data have some characteristics:the structure changes over time and the coefficient of data is varying, the characteristics can not meet the precondition of the classic linear models. So the classic linear models can neither reflect the changes of data in the micro nor accurately fit and resolve the time-series data. The varying-coefficient models constructed by state-space model introduce the conception of "state". The varying-coefficient models can dynamically reflect the changes in micro and accurately fit and forecast the data. Trough applying the varying-coefficient models in time-series data and comparing with the constant-coefficient models, this paper reveal the advantage of the varying-coefficient models and provide new methods and ideas for following study about time-series data.The time-series/cross data contains three dimensions information:time, cross and index. The classic linear models only take use of two dimensions information. Trough using varying-intercept models to fit the time-series/cross data, this paper introduce a new method and ideal for following study.Methods:The principle and methods of varying-parameters models will be introduced in this paper. According to the data that was published byand, this paper constructed two models:the varying-coefficient models about the relationship between 1978-2007 national per capita income and per capita spending on health, and the varying-intercept models about around 2002-2008 per capita annual consumption expenditure and per capita health care spending, and revealing the advantage of the varying-parameters models. This paper takes use of Eviews5.0 to analysis data.Results:The fitting results of varying-coefficient models shows that the coefficient between per capita income and per capita health spending is varying with time.And the residual sum of squares of varying-coefficient models about per capita income and per capita health spending is smaller than the constant-coefficient.Through comprehensively analyzing the per capita income time-series data and the regional income disparities by the varying-intercept models, we can get the results that in China the medical spontaneous propensity to consume of eastern state is significantly higher than the western state. And the medical spontaneous propensity of Beijing is the highest, while Jiangxi is the lowest. During 2002-2008 the Chinese medical spontaneous propensity to consume is gradually increasing.Conclusion:The varying-coefficient models have a great advantage in fitting and resolving the large span time time-series data compare to the constant-coefficient models.The varying-intercept models can comprehensively analysis the information of three-dimension data, and the varying-intercept has the applied advantage compared to the classical linear models.