Karpusha M. Modeling and identification problems of portfolio optimization and risk management. -

A new discrete continuous model with double-digit and three-digit dummies variables was introduced. These models have the following features. They involve a preliminary analysis of data on stationary and nonstationarity. The pretreatment procedures are done for income data, splitting time series into classes and classifying them (TS or DS class). The TS class includes series that are stationary relatively to the deterministic trend, so the trend component is allocated for them. The DS class consists of rows with stochastic trend being present (possibly with a deterministic trend). These numbers are reduced to fixed by k-fold differentiation. These two classes require to be simulated with different methods. Explore different effects availability in the time series data. The econometric modeling was done in respect to the fact that both continuous and discrete effects could be observed. Also, neither time moments, when regresants values bounces nor magnitude of these bounces are known. Therefore, the model and the procedure of it is construction is such that the parameters of the model and discrete effects are identified by iteration. The interaction of continuous and discrete components adjusts regresant to adequate reflection of the real process. Forecasting models use logit-models and probit-models to determine the most likely option in price changing and to get high predictive properties.