Application Of Markov Mechanism To Change Mixed Frequency Data Model

Macroeconomic time series is an important macroeconomic indicator reaction. There are many macroeconomic time series data to reflect the current economic situation, and sometimes even as a basis for future economic trends make useful predictions, such as quarter GDP (gross domestic product) data, the monthly CPI (consumer price index) data, PPI (producer price index) data. This is where the data will have a direct or indirect economic trends influence, which attracted widespread attention individual economic, business and national and international community, for these numerous and complex data, people try to use different data processing methods, establish each Species grab information from the model to achieve the purpose, and this information will serve as a model projections and estimates, on everyday savings, investment behavior and major decision-making provide reference and basis for economic behavior. Time series data is complex, different lengths, different sampling frequency, data attributes of different features.Time series models are usually built on the same frequency data. Under different circumstances the data frequency, sometimes using the sum or an alternative way to become the low frequency of the high frequency data processing data, the same, if the low-frequency data processed into high-frequency data, it is necessary to insert a high-frequency data in the low frequency data, also known as the insertion method, which through the same frequency of data can be processed using the traditional time series model for analysis. However, the high frequency data aggregated into a low-frequency data, a lot of high-frequency data sample information is ignored, then the volatility attribute the high frequency data will disappear, it can be said to reduce the total sample plus artificial way information capacity; the corresponding value in the low-frequency data is inserted, the high frequency values are not accurate data on behalf of the entity economy, resulting in that the structure of the traces, now mostly interpolation method is a method of pure mathematics, the data is complete, but the lack of real sex, lack of theoretical basis. In practice, in the economic field, the high frequency data processing for the low frequency data are used in many studies among the total employed are added or alternative methods, there are a small part of the data required in order to study the model, the lack of data too and it will handle the low-frequency data into a high-frequency data. But this approach is not only cumbersome and less reliable conclusions, so if you can be a high-frequency data of different frequencies and low frequency data directly into a model, the above problems will be solved, based on this need and circumstances, resulting in a mixed frequency data model.Mixed frequency data model compared to the traditional data model has high prediction accuracy and timeliness of advanced predictive characteristics, combined with other advanced metering method can not only improve the accuracy of model estimates, can also improve traditional data analysis model. On this basis, if we study the frequency of the data is inconsistent, a hybrid data model will be able to do more with less frequency.This article is structured as follows:The first part:Ⅰ introduce the mix frequency model including MIDAS model,CoMIDAS model, FaMIDAS model.Part Ⅱ:I combined FaMIDAS model with Markov switching model. And then I got the Markov switching factor mixed frequency data sampling model (MS-FaMIDAS model). And I give the method of estimation and test of the MS-FaMIDAS model.Part Ⅲ:According to the research conclusions and achievements, I gives some research suggestions and research directions.